* [PATCH v3 0/21] pack bitmaps
@ 2013-11-14 12:41 Jeff King
2013-11-14 12:42 ` [PATCH v3 01/21] sha1write: make buffer const-correct Jeff King
` (21 more replies)
0 siblings, 22 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:41 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
Here's another iteration of the pack bitmaps series. Compared to v2, it
changes:
- misc style/typo fixes
- portability fixes from Ramsay and Torsten
- count-objects garbage-reporting patch from Duy
- disable bitmaps when is_repository_shallow(); this also covers the
case where the client is shallow, since we feed pack-objects a
--shallow-file in that case. This used to done by checking
!internal_rev_list, but that doesn't apply after cdab485.
- ewah sources now properly use git-compat-util.h and do not include
system headers
- the ewah code uses ewah_malloc, ewah_realloc, and so forth to let the
project use a particular allocator (and we want to use xmalloc and
friends). And we defined those in pack-bitmap.h, but of course that
had no effect on the ewah/*.c files that did not include
pack-bitmap.h. Since we are hacking up and git-ifying libewok
anyway, we can just set the hardcoded fallback to xmalloc instead of
malloc.
- the ewah code used gcc's __builtin_ctzll, but did not provide a
suitable fallback. We now provide a fallback in C.
- The bitmap reading code only handles a single bitmapped pack (since
they must be fully closed, there is not much point in having
multiple). It used to silently ignore extra bitmap indices it found,
but will now warn that they are being ignored.
- The name-hash cache is now optional, controlled by
pack.writeBitmapHashCache.
- The test script will now do basic interoperability testing with jgit
(if you have jgit in your $PATH).
- There are now perf tests. Spoiler alert: bitmaps make clones faster.
See patch 20 for details. We can also measure the speedup from the
hash cache (see patch 21).
Not addressed:
- I did not include the NEEDS_ALIGNED_ACCESS patch. I note that we do
not even have a Makefile knob for this, and the code in read-cache.c
has probably never actually been used. Are there real systems that
have a problem? The read-cache code was in support of the index v4
experiment, which did away with the 8-byte padding. So it could be
that we simply don't see it, because everything is currently
aligned.
- On a related note, we do some cast-buffer-to-struct magic on the
mmap'd file. I note that the regular packfile reader also does this.
How careful do we want to be?
- We still assume that reusing a slice from the front of the pack will
never miss delta bases. This is the case currently for packs
generated by both git and JGit, but it would be nice to mark the
property in the bitmap index. Adding a new flag would break JGit
compatibility, though. We can either make it an option, or assume
it's good enough for now and worry about it in v2.
[01/21]: sha1write: make buffer const-correct
[02/21]: revindex: Export new APIs
[03/21]: pack-objects: Refactor the packing list
[04/21]: pack-objects: factor out name_hash
[05/21]: revision: allow setting custom limiter function
[06/21]: sha1_file: export `git_open_noatime`
[07/21]: compat: add endianness helpers
[08/21]: ewah: compressed bitmap implementation
[09/21]: documentation: add documentation for the bitmap format
[10/21]: pack-bitmap: add support for bitmap indexes
[11/21]: pack-objects: use bitmaps when packing objects
[12/21]: rev-list: add bitmap mode to speed up object lists
[13/21]: pack-objects: implement bitmap writing
[14/21]: repack: stop using magic number for ARRAY_SIZE(exts)
[15/21]: repack: turn exts array into array-of-struct
[16/21]: repack: handle optional files created by pack-objects
[17/21]: repack: consider bitmaps when performing repacks
[18/21]: count-objects: recognize .bitmap in garbage-checking
[19/21]: t: add basic bitmap functionality tests
[20/21]: t/perf: add tests for pack bitmaps
[21/21]: pack-bitmap: implement optional name_hash cache
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* [PATCH v3 01/21] sha1write: make buffer const-correct
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
@ 2013-11-14 12:42 ` Jeff King
2013-11-14 12:42 ` [PATCH v3 02/21] revindex: Export new APIs Jeff King
` (20 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:42 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
We are passed a "void *" and write it out without ever
touching it; let's indicate that by using "const".
Signed-off-by: Jeff King <peff@peff.net>
---
csum-file.c | 6 +++---
csum-file.h | 2 +-
2 files changed, 4 insertions(+), 4 deletions(-)
diff --git a/csum-file.c b/csum-file.c
index 53f5375..465971c 100644
--- a/csum-file.c
+++ b/csum-file.c
@@ -11,7 +11,7 @@
#include "progress.h"
#include "csum-file.h"
-static void flush(struct sha1file *f, void *buf, unsigned int count)
+static void flush(struct sha1file *f, const void *buf, unsigned int count)
{
if (0 <= f->check_fd && count) {
unsigned char check_buffer[8192];
@@ -86,13 +86,13 @@ int sha1close(struct sha1file *f, unsigned char *result, unsigned int flags)
return fd;
}
-int sha1write(struct sha1file *f, void *buf, unsigned int count)
+int sha1write(struct sha1file *f, const void *buf, unsigned int count)
{
while (count) {
unsigned offset = f->offset;
unsigned left = sizeof(f->buffer) - offset;
unsigned nr = count > left ? left : count;
- void *data;
+ const void *data;
if (f->do_crc)
f->crc32 = crc32(f->crc32, buf, nr);
diff --git a/csum-file.h b/csum-file.h
index 3b540bd..9dedb03 100644
--- a/csum-file.h
+++ b/csum-file.h
@@ -34,7 +34,7 @@ extern struct sha1file *sha1fd(int fd, const char *name);
extern struct sha1file *sha1fd_check(const char *name);
extern struct sha1file *sha1fd_throughput(int fd, const char *name, struct progress *tp);
extern int sha1close(struct sha1file *, unsigned char *, unsigned int);
-extern int sha1write(struct sha1file *, void *, unsigned int);
+extern int sha1write(struct sha1file *, const void *, unsigned int);
extern void sha1flush(struct sha1file *f);
extern void crc32_begin(struct sha1file *);
extern uint32_t crc32_end(struct sha1file *);
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 02/21] revindex: Export new APIs
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
2013-11-14 12:42 ` [PATCH v3 01/21] sha1write: make buffer const-correct Jeff King
@ 2013-11-14 12:42 ` Jeff King
2013-11-14 12:42 ` [PATCH v3 03/21] pack-objects: Refactor the packing list Jeff King
` (19 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:42 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
Allow users to efficiently lookup consecutive entries that are expected
to be found on the same revindex by exporting `find_revindex_position`:
this function takes a pointer to revindex itself, instead of looking up
the proper revindex for a given packfile on each call.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
pack-revindex.c | 38 +++++++++++++++++++++++++-------------
pack-revindex.h | 8 ++++++++
2 files changed, 33 insertions(+), 13 deletions(-)
diff --git a/pack-revindex.c b/pack-revindex.c
index b4d2b35..0bb13b1 100644
--- a/pack-revindex.c
+++ b/pack-revindex.c
@@ -16,11 +16,6 @@
* get the object sha1 from the main index.
*/
-struct pack_revindex {
- struct packed_git *p;
- struct revindex_entry *revindex;
-};
-
static struct pack_revindex *pack_revindex;
static int pack_revindex_hashsz;
@@ -201,15 +196,14 @@ static void create_pack_revindex(struct pack_revindex *rix)
sort_revindex(rix->revindex, num_ent, p->pack_size);
}
-struct revindex_entry *find_pack_revindex(struct packed_git *p, off_t ofs)
+struct pack_revindex *revindex_for_pack(struct packed_git *p)
{
int num;
- unsigned lo, hi;
struct pack_revindex *rix;
- struct revindex_entry *revindex;
if (!pack_revindex_hashsz)
init_pack_revindex();
+
num = pack_revindex_ix(p);
if (num < 0)
die("internal error: pack revindex fubar");
@@ -217,21 +211,39 @@ struct revindex_entry *find_pack_revindex(struct packed_git *p, off_t ofs)
rix = &pack_revindex[num];
if (!rix->revindex)
create_pack_revindex(rix);
- revindex = rix->revindex;
- lo = 0;
- hi = p->num_objects + 1;
+ return rix;
+}
+
+int find_revindex_position(struct pack_revindex *pridx, off_t ofs)
+{
+ int lo = 0;
+ int hi = pridx->p->num_objects + 1;
+ struct revindex_entry *revindex = pridx->revindex;
+
do {
unsigned mi = lo + (hi - lo) / 2;
if (revindex[mi].offset == ofs) {
- return revindex + mi;
+ return mi;
} else if (ofs < revindex[mi].offset)
hi = mi;
else
lo = mi + 1;
} while (lo < hi);
+
error("bad offset for revindex");
- return NULL;
+ return -1;
+}
+
+struct revindex_entry *find_pack_revindex(struct packed_git *p, off_t ofs)
+{
+ struct pack_revindex *pridx = revindex_for_pack(p);
+ int pos = find_revindex_position(pridx, ofs);
+
+ if (pos < 0)
+ return NULL;
+
+ return pridx->revindex + pos;
}
void discard_revindex(void)
diff --git a/pack-revindex.h b/pack-revindex.h
index 8d5027a..866ca9c 100644
--- a/pack-revindex.h
+++ b/pack-revindex.h
@@ -6,6 +6,14 @@ struct revindex_entry {
unsigned int nr;
};
+struct pack_revindex {
+ struct packed_git *p;
+ struct revindex_entry *revindex;
+};
+
+struct pack_revindex *revindex_for_pack(struct packed_git *p);
+int find_revindex_position(struct pack_revindex *pridx, off_t ofs);
+
struct revindex_entry *find_pack_revindex(struct packed_git *p, off_t ofs);
void discard_revindex(void);
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 03/21] pack-objects: Refactor the packing list
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
2013-11-14 12:42 ` [PATCH v3 01/21] sha1write: make buffer const-correct Jeff King
2013-11-14 12:42 ` [PATCH v3 02/21] revindex: Export new APIs Jeff King
@ 2013-11-14 12:42 ` Jeff King
2013-11-14 12:43 ` [PATCH v3 04/21] pack-objects: factor out name_hash Jeff King
` (18 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:42 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
The hash table that stores the packing list for a given `pack-objects`
run was tightly coupled to the pack-objects code.
In this commit, we refactor the hash table and the underlying storage
array into a `packing_data` struct. The functionality for accessing and
adding entries to the packing list is hence accessible from other parts
of Git besides the `pack-objects` builtin.
This refactoring is a requirement for further patches in this series
that will require accessing the commit packing list from outside of
`pack-objects`.
The hash table implementation has been minimally altered: we now
use table sizes which are always a power of two, to ensure a uniform
index distribution in the array.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Makefile | 2 +
builtin/pack-objects.c | 175 +++++++++++--------------------------------------
pack-objects.c | 111 +++++++++++++++++++++++++++++++
pack-objects.h | 47 +++++++++++++
4 files changed, 200 insertions(+), 135 deletions(-)
create mode 100644 pack-objects.c
create mode 100644 pack-objects.h
diff --git a/Makefile b/Makefile
index af847f8..48ff0bd 100644
--- a/Makefile
+++ b/Makefile
@@ -694,6 +694,7 @@ LIB_H += notes-merge.h
LIB_H += notes-utils.h
LIB_H += notes.h
LIB_H += object.h
+LIB_H += pack-objects.h
LIB_H += pack-revindex.h
LIB_H += pack.h
LIB_H += parse-options.h
@@ -831,6 +832,7 @@ LIB_OBJS += notes-merge.o
LIB_OBJS += notes-utils.o
LIB_OBJS += object.o
LIB_OBJS += pack-check.o
+LIB_OBJS += pack-objects.o
LIB_OBJS += pack-revindex.o
LIB_OBJS += pack-write.o
LIB_OBJS += pager.o
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 36273dd..f3f0cf9 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -14,6 +14,7 @@
#include "diff.h"
#include "revision.h"
#include "list-objects.h"
+#include "pack-objects.h"
#include "progress.h"
#include "refs.h"
#include "streaming.h"
@@ -25,42 +26,15 @@ static const char *pack_usage[] = {
NULL
};
-struct object_entry {
- struct pack_idx_entry idx;
- unsigned long size; /* uncompressed size */
- struct packed_git *in_pack; /* already in pack */
- off_t in_pack_offset;
- struct object_entry *delta; /* delta base object */
- struct object_entry *delta_child; /* deltified objects who bases me */
- struct object_entry *delta_sibling; /* other deltified objects who
- * uses the same base as me
- */
- void *delta_data; /* cached delta (uncompressed) */
- unsigned long delta_size; /* delta data size (uncompressed) */
- unsigned long z_delta_size; /* delta data size (compressed) */
- enum object_type type;
- enum object_type in_pack_type; /* could be delta */
- uint32_t hash; /* name hint hash */
- unsigned char in_pack_header_size;
- unsigned preferred_base:1; /*
- * we do not pack this, but is available
- * to be used as the base object to delta
- * objects against.
- */
- unsigned no_try_delta:1;
- unsigned tagged:1; /* near the very tip of refs */
- unsigned filled:1; /* assigned write-order */
-};
-
/*
- * Objects we are going to pack are collected in objects array (dynamically
- * expanded). nr_objects & nr_alloc controls this array. They are stored
- * in the order we see -- typically rev-list --objects order that gives us
- * nice "minimum seek" order.
+ * Objects we are going to pack are collected in the `to_pack` structure.
+ * It contains an array (dynamically expanded) of the object data, and a map
+ * that can resolve SHA1s to their position in the array.
*/
-static struct object_entry *objects;
+static struct packing_data to_pack;
+
static struct pack_idx_entry **written_list;
-static uint32_t nr_objects, nr_alloc, nr_result, nr_written;
+static uint32_t nr_result, nr_written;
static int non_empty;
static int reuse_delta = 1, reuse_object = 1;
@@ -90,21 +64,11 @@ static unsigned long cache_max_small_delta_size = 1000;
static unsigned long window_memory_limit = 0;
/*
- * The object names in objects array are hashed with this hashtable,
- * to help looking up the entry by object name.
- * This hashtable is built after all the objects are seen.
- */
-static int *object_ix;
-static int object_ix_hashsz;
-static struct object_entry *locate_object_entry(const unsigned char *sha1);
-
-/*
* stats
*/
static uint32_t written, written_delta;
static uint32_t reused, reused_delta;
-
static void *get_delta(struct object_entry *entry)
{
unsigned long size, base_size, delta_size;
@@ -553,12 +517,12 @@ static int mark_tagged(const char *path, const unsigned char *sha1, int flag,
void *cb_data)
{
unsigned char peeled[20];
- struct object_entry *entry = locate_object_entry(sha1);
+ struct object_entry *entry = packlist_find(&to_pack, sha1, NULL);
if (entry)
entry->tagged = 1;
if (!peel_ref(path, peeled)) {
- entry = locate_object_entry(peeled);
+ entry = packlist_find(&to_pack, peeled, NULL);
if (entry)
entry->tagged = 1;
}
@@ -633,9 +597,10 @@ static struct object_entry **compute_write_order(void)
{
unsigned int i, wo_end, last_untagged;
- struct object_entry **wo = xmalloc(nr_objects * sizeof(*wo));
+ struct object_entry **wo = xmalloc(to_pack.nr_objects * sizeof(*wo));
+ struct object_entry *objects = to_pack.objects;
- for (i = 0; i < nr_objects; i++) {
+ for (i = 0; i < to_pack.nr_objects; i++) {
objects[i].tagged = 0;
objects[i].filled = 0;
objects[i].delta_child = NULL;
@@ -647,7 +612,7 @@ static struct object_entry **compute_write_order(void)
* Make sure delta_sibling is sorted in the original
* recency order.
*/
- for (i = nr_objects; i > 0;) {
+ for (i = to_pack.nr_objects; i > 0;) {
struct object_entry *e = &objects[--i];
if (!e->delta)
continue;
@@ -665,7 +630,7 @@ static struct object_entry **compute_write_order(void)
* Give the objects in the original recency order until
* we see a tagged tip.
*/
- for (i = wo_end = 0; i < nr_objects; i++) {
+ for (i = wo_end = 0; i < to_pack.nr_objects; i++) {
if (objects[i].tagged)
break;
add_to_write_order(wo, &wo_end, &objects[i]);
@@ -675,7 +640,7 @@ static struct object_entry **compute_write_order(void)
/*
* Then fill all the tagged tips.
*/
- for (; i < nr_objects; i++) {
+ for (; i < to_pack.nr_objects; i++) {
if (objects[i].tagged)
add_to_write_order(wo, &wo_end, &objects[i]);
}
@@ -683,7 +648,7 @@ static struct object_entry **compute_write_order(void)
/*
* And then all remaining commits and tags.
*/
- for (i = last_untagged; i < nr_objects; i++) {
+ for (i = last_untagged; i < to_pack.nr_objects; i++) {
if (objects[i].type != OBJ_COMMIT &&
objects[i].type != OBJ_TAG)
continue;
@@ -693,7 +658,7 @@ static struct object_entry **compute_write_order(void)
/*
* And then all the trees.
*/
- for (i = last_untagged; i < nr_objects; i++) {
+ for (i = last_untagged; i < to_pack.nr_objects; i++) {
if (objects[i].type != OBJ_TREE)
continue;
add_to_write_order(wo, &wo_end, &objects[i]);
@@ -702,13 +667,13 @@ static struct object_entry **compute_write_order(void)
/*
* Finally all the rest in really tight order
*/
- for (i = last_untagged; i < nr_objects; i++) {
+ for (i = last_untagged; i < to_pack.nr_objects; i++) {
if (!objects[i].filled)
add_family_to_write_order(wo, &wo_end, &objects[i]);
}
- if (wo_end != nr_objects)
- die("ordered %u objects, expected %"PRIu32, wo_end, nr_objects);
+ if (wo_end != to_pack.nr_objects)
+ die("ordered %u objects, expected %"PRIu32, wo_end, to_pack.nr_objects);
return wo;
}
@@ -724,7 +689,7 @@ static void write_pack_file(void)
if (progress > pack_to_stdout)
progress_state = start_progress("Writing objects", nr_result);
- written_list = xmalloc(nr_objects * sizeof(*written_list));
+ written_list = xmalloc(to_pack.nr_objects * sizeof(*written_list));
write_order = compute_write_order();
do {
@@ -740,7 +705,7 @@ static void write_pack_file(void)
if (!offset)
die_errno("unable to write pack header");
nr_written = 0;
- for (; i < nr_objects; i++) {
+ for (; i < to_pack.nr_objects; i++) {
struct object_entry *e = write_order[i];
if (write_one(f, e, &offset) == WRITE_ONE_BREAK)
break;
@@ -803,7 +768,7 @@ static void write_pack_file(void)
written_list[j]->offset = (off_t)-1;
}
nr_remaining -= nr_written;
- } while (nr_remaining && i < nr_objects);
+ } while (nr_remaining && i < to_pack.nr_objects);
free(written_list);
free(write_order);
@@ -813,53 +778,6 @@ static void write_pack_file(void)
written, nr_result);
}
-static int locate_object_entry_hash(const unsigned char *sha1)
-{
- int i;
- unsigned int ui;
- memcpy(&ui, sha1, sizeof(unsigned int));
- i = ui % object_ix_hashsz;
- while (0 < object_ix[i]) {
- if (!hashcmp(sha1, objects[object_ix[i] - 1].idx.sha1))
- return i;
- if (++i == object_ix_hashsz)
- i = 0;
- }
- return -1 - i;
-}
-
-static struct object_entry *locate_object_entry(const unsigned char *sha1)
-{
- int i;
-
- if (!object_ix_hashsz)
- return NULL;
-
- i = locate_object_entry_hash(sha1);
- if (0 <= i)
- return &objects[object_ix[i]-1];
- return NULL;
-}
-
-static void rehash_objects(void)
-{
- uint32_t i;
- struct object_entry *oe;
-
- object_ix_hashsz = nr_objects * 3;
- if (object_ix_hashsz < 1024)
- object_ix_hashsz = 1024;
- object_ix = xrealloc(object_ix, sizeof(int) * object_ix_hashsz);
- memset(object_ix, 0, sizeof(int) * object_ix_hashsz);
- for (i = 0, oe = objects; i < nr_objects; i++, oe++) {
- int ix = locate_object_entry_hash(oe->idx.sha1);
- if (0 <= ix)
- continue;
- ix = -1 - ix;
- object_ix[ix] = i + 1;
- }
-}
-
static uint32_t name_hash(const char *name)
{
uint32_t c, hash = 0;
@@ -908,13 +826,12 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
struct object_entry *entry;
struct packed_git *p, *found_pack = NULL;
off_t found_offset = 0;
- int ix;
uint32_t hash = name_hash(name);
+ uint32_t index_pos;
- ix = nr_objects ? locate_object_entry_hash(sha1) : -1;
- if (ix >= 0) {
+ entry = packlist_find(&to_pack, sha1, &index_pos);
+ if (entry) {
if (exclude) {
- entry = objects + object_ix[ix] - 1;
if (!entry->preferred_base)
nr_result--;
entry->preferred_base = 1;
@@ -947,14 +864,7 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
}
}
- if (nr_objects >= nr_alloc) {
- nr_alloc = (nr_alloc + 1024) * 3 / 2;
- objects = xrealloc(objects, nr_alloc * sizeof(*entry));
- }
-
- entry = objects + nr_objects++;
- memset(entry, 0, sizeof(*entry));
- hashcpy(entry->idx.sha1, sha1);
+ entry = packlist_alloc(&to_pack, sha1, index_pos);
entry->hash = hash;
if (type)
entry->type = type;
@@ -967,12 +877,7 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
entry->in_pack_offset = found_offset;
}
- if (object_ix_hashsz * 3 <= nr_objects * 4)
- rehash_objects();
- else
- object_ix[-1 - ix] = nr_objects;
-
- display_progress(progress_state, nr_objects);
+ display_progress(progress_state, to_pack.nr_objects);
if (name && no_try_delta(name))
entry->no_try_delta = 1;
@@ -1329,7 +1234,7 @@ static void check_object(struct object_entry *entry)
break;
}
- if (base_ref && (base_entry = locate_object_entry(base_ref))) {
+ if (base_ref && (base_entry = packlist_find(&to_pack, base_ref, NULL))) {
/*
* If base_ref was set above that means we wish to
* reuse delta data, and we even found that base
@@ -1403,12 +1308,12 @@ static void get_object_details(void)
uint32_t i;
struct object_entry **sorted_by_offset;
- sorted_by_offset = xcalloc(nr_objects, sizeof(struct object_entry *));
- for (i = 0; i < nr_objects; i++)
- sorted_by_offset[i] = objects + i;
- qsort(sorted_by_offset, nr_objects, sizeof(*sorted_by_offset), pack_offset_sort);
+ sorted_by_offset = xcalloc(to_pack.nr_objects, sizeof(struct object_entry *));
+ for (i = 0; i < to_pack.nr_objects; i++)
+ sorted_by_offset[i] = to_pack.objects + i;
+ qsort(sorted_by_offset, to_pack.nr_objects, sizeof(*sorted_by_offset), pack_offset_sort);
- for (i = 0; i < nr_objects; i++) {
+ for (i = 0; i < to_pack.nr_objects; i++) {
struct object_entry *entry = sorted_by_offset[i];
check_object(entry);
if (big_file_threshold < entry->size)
@@ -2034,7 +1939,7 @@ static int add_ref_tag(const char *path, const unsigned char *sha1, int flag, vo
if (!prefixcmp(path, "refs/tags/") && /* is a tag? */
!peel_ref(path, peeled) && /* peelable? */
- locate_object_entry(peeled)) /* object packed? */
+ packlist_find(&to_pack, peeled, NULL)) /* object packed? */
add_object_entry(sha1, OBJ_TAG, NULL, 0);
return 0;
}
@@ -2057,14 +1962,14 @@ static void prepare_pack(int window, int depth)
if (!pack_to_stdout)
do_check_packed_object_crc = 1;
- if (!nr_objects || !window || !depth)
+ if (!to_pack.nr_objects || !window || !depth)
return;
- delta_list = xmalloc(nr_objects * sizeof(*delta_list));
+ delta_list = xmalloc(to_pack.nr_objects * sizeof(*delta_list));
nr_deltas = n = 0;
- for (i = 0; i < nr_objects; i++) {
- struct object_entry *entry = objects + i;
+ for (i = 0; i < to_pack.nr_objects; i++) {
+ struct object_entry *entry = to_pack.objects + i;
if (entry->delta)
/* This happens if we decided to reuse existing
@@ -2342,7 +2247,7 @@ static void loosen_unused_packed_objects(struct rev_info *revs)
for (i = 0; i < p->num_objects; i++) {
sha1 = nth_packed_object_sha1(p, i);
- if (!locate_object_entry(sha1) &&
+ if (!packlist_find(&to_pack, sha1, NULL) &&
!has_sha1_pack_kept_or_nonlocal(sha1))
if (force_object_loose(sha1, p->mtime))
die("unable to force loose object");
diff --git a/pack-objects.c b/pack-objects.c
new file mode 100644
index 0000000..d01d851
--- /dev/null
+++ b/pack-objects.c
@@ -0,0 +1,111 @@
+#include "cache.h"
+#include "object.h"
+#include "pack.h"
+#include "pack-objects.h"
+
+static uint32_t locate_object_entry_hash(struct packing_data *pdata,
+ const unsigned char *sha1,
+ int *found)
+{
+ uint32_t i, hash, mask = (pdata->index_size - 1);
+
+ memcpy(&hash, sha1, sizeof(uint32_t));
+ i = hash & mask;
+
+ while (pdata->index[i] > 0) {
+ uint32_t pos = pdata->index[i] - 1;
+
+ if (!hashcmp(sha1, pdata->objects[pos].idx.sha1)) {
+ *found = 1;
+ return i;
+ }
+
+ i = (i + 1) & mask;
+ }
+
+ *found = 0;
+ return i;
+}
+
+static inline uint32_t closest_pow2(uint32_t v)
+{
+ v = v - 1;
+ v |= v >> 1;
+ v |= v >> 2;
+ v |= v >> 4;
+ v |= v >> 8;
+ v |= v >> 16;
+ return v + 1;
+}
+
+static void rehash_objects(struct packing_data *pdata)
+{
+ uint32_t i;
+ struct object_entry *entry;
+
+ pdata->index_size = closest_pow2(pdata->nr_objects * 3);
+ if (pdata->index_size < 1024)
+ pdata->index_size = 1024;
+
+ pdata->index = xrealloc(pdata->index, sizeof(uint32_t) * pdata->index_size);
+ memset(pdata->index, 0, sizeof(int) * pdata->index_size);
+
+ entry = pdata->objects;
+
+ for (i = 0; i < pdata->nr_objects; i++) {
+ int found;
+ uint32_t ix = locate_object_entry_hash(pdata, entry->idx.sha1, &found);
+
+ if (found)
+ die("BUG: Duplicate object in hash");
+
+ pdata->index[ix] = i + 1;
+ entry++;
+ }
+}
+
+struct object_entry *packlist_find(struct packing_data *pdata,
+ const unsigned char *sha1,
+ uint32_t *index_pos)
+{
+ uint32_t i;
+ int found;
+
+ if (!pdata->index_size)
+ return NULL;
+
+ i = locate_object_entry_hash(pdata, sha1, &found);
+
+ if (index_pos)
+ *index_pos = i;
+
+ if (!found)
+ return NULL;
+
+ return &pdata->objects[pdata->index[i] - 1];
+}
+
+struct object_entry *packlist_alloc(struct packing_data *pdata,
+ const unsigned char *sha1,
+ uint32_t index_pos)
+{
+ struct object_entry *new_entry;
+
+ if (pdata->nr_objects >= pdata->nr_alloc) {
+ pdata->nr_alloc = (pdata->nr_alloc + 1024) * 3 / 2;
+ pdata->objects = xrealloc(pdata->objects,
+ pdata->nr_alloc * sizeof(*new_entry));
+ }
+
+ new_entry = pdata->objects + pdata->nr_objects++;
+
+ memset(new_entry, 0, sizeof(*new_entry));
+ hashcpy(new_entry->idx.sha1, sha1);
+
+ if (pdata->index_size * 3 <= pdata->nr_objects * 4)
+ rehash_objects(pdata);
+ else
+ pdata->index[index_pos] = pdata->nr_objects;
+
+ return new_entry;
+}
diff --git a/pack-objects.h b/pack-objects.h
new file mode 100644
index 0000000..f528215
--- /dev/null
+++ b/pack-objects.h
@@ -0,0 +1,47 @@
+#ifndef PACK_OBJECTS_H
+#define PACK_OBJECTS_H
+
+struct object_entry {
+ struct pack_idx_entry idx;
+ unsigned long size; /* uncompressed size */
+ struct packed_git *in_pack; /* already in pack */
+ off_t in_pack_offset;
+ struct object_entry *delta; /* delta base object */
+ struct object_entry *delta_child; /* deltified objects who bases me */
+ struct object_entry *delta_sibling; /* other deltified objects who
+ * uses the same base as me
+ */
+ void *delta_data; /* cached delta (uncompressed) */
+ unsigned long delta_size; /* delta data size (uncompressed) */
+ unsigned long z_delta_size; /* delta data size (compressed) */
+ enum object_type type;
+ enum object_type in_pack_type; /* could be delta */
+ uint32_t hash; /* name hint hash */
+ unsigned char in_pack_header_size;
+ unsigned preferred_base:1; /*
+ * we do not pack this, but is available
+ * to be used as the base object to delta
+ * objects against.
+ */
+ unsigned no_try_delta:1;
+ unsigned tagged:1; /* near the very tip of refs */
+ unsigned filled:1; /* assigned write-order */
+};
+
+struct packing_data {
+ struct object_entry *objects;
+ uint32_t nr_objects, nr_alloc;
+
+ int32_t *index;
+ uint32_t index_size;
+};
+
+struct object_entry *packlist_alloc(struct packing_data *pdata,
+ const unsigned char *sha1,
+ uint32_t index_pos);
+
+struct object_entry *packlist_find(struct packing_data *pdata,
+ const unsigned char *sha1,
+ uint32_t *index_pos);
+
+#endif
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 04/21] pack-objects: factor out name_hash
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (2 preceding siblings ...)
2013-11-14 12:42 ` [PATCH v3 03/21] pack-objects: Refactor the packing list Jeff King
@ 2013-11-14 12:43 ` Jeff King
2013-11-14 12:43 ` [PATCH v3 05/21] revision: allow setting custom limiter function Jeff King
` (17 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:43 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
As the pack-objects system grows beyond the single
pack-objects.c file, more parts (like the soon-to-exist
bitmap code) will need to compute hashes for matching
deltas. Factor out name_hash to make it available to other
files.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
builtin/pack-objects.c | 24 ++----------------------
pack-objects.h | 20 ++++++++++++++++++++
2 files changed, 22 insertions(+), 22 deletions(-)
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index f3f0cf9..faf746b 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -778,26 +778,6 @@ static void write_pack_file(void)
written, nr_result);
}
-static uint32_t name_hash(const char *name)
-{
- uint32_t c, hash = 0;
-
- if (!name)
- return 0;
-
- /*
- * This effectively just creates a sortable number from the
- * last sixteen non-whitespace characters. Last characters
- * count "most", so things that end in ".c" sort together.
- */
- while ((c = *name++) != 0) {
- if (isspace(c))
- continue;
- hash = (hash >> 2) + (c << 24);
- }
- return hash;
-}
-
static void setup_delta_attr_check(struct git_attr_check *check)
{
static struct git_attr *attr_delta;
@@ -826,7 +806,7 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
struct object_entry *entry;
struct packed_git *p, *found_pack = NULL;
off_t found_offset = 0;
- uint32_t hash = name_hash(name);
+ uint32_t hash = pack_name_hash(name);
uint32_t index_pos;
entry = packlist_find(&to_pack, sha1, &index_pos);
@@ -1082,7 +1062,7 @@ static void add_preferred_base_object(const char *name)
{
struct pbase_tree *it;
int cmplen;
- unsigned hash = name_hash(name);
+ unsigned hash = pack_name_hash(name);
if (!num_preferred_base || check_pbase_path(hash))
return;
diff --git a/pack-objects.h b/pack-objects.h
index f528215..90ad0a8 100644
--- a/pack-objects.h
+++ b/pack-objects.h
@@ -44,4 +44,24 @@ struct object_entry *packlist_find(struct packing_data *pdata,
const unsigned char *sha1,
uint32_t *index_pos);
+static inline uint32_t pack_name_hash(const char *name)
+{
+ uint32_t c, hash = 0;
+
+ if (!name)
+ return 0;
+
+ /*
+ * This effectively just creates a sortable number from the
+ * last sixteen non-whitespace characters. Last characters
+ * count "most", so things that end in ".c" sort together.
+ */
+ while ((c = *name++) != 0) {
+ if (isspace(c))
+ continue;
+ hash = (hash >> 2) + (c << 24);
+ }
+ return hash;
+}
+
#endif
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 05/21] revision: allow setting custom limiter function
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (3 preceding siblings ...)
2013-11-14 12:43 ` [PATCH v3 04/21] pack-objects: factor out name_hash Jeff King
@ 2013-11-14 12:43 ` Jeff King
2013-11-14 12:43 ` [PATCH v3 06/21] sha1_file: export `git_open_noatime` Jeff King
` (16 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:43 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
This commit enables users of `struct rev_info` to peform custom limiting
during a revision walk (i.e. `get_revision`).
If the field `include_check` has been set to a callback, this callback
will be issued once for each commit before it is added to the "pending"
list of the revwalk. If the include check returns 0, the commit will be
marked as added but won't be pushed to the pending list, effectively
limiting the walk.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
revision.c | 4 ++++
revision.h | 2 ++
2 files changed, 6 insertions(+)
diff --git a/revision.c b/revision.c
index 956040c..260f4a1 100644
--- a/revision.c
+++ b/revision.c
@@ -779,6 +779,10 @@ static int add_parents_to_list(struct rev_info *revs, struct commit *commit,
return 0;
commit->object.flags |= ADDED;
+ if (revs->include_check &&
+ !revs->include_check(commit, revs->include_check_data))
+ return 0;
+
/*
* If the commit is uninteresting, don't try to
* prune parents - we want the maximal uninteresting
diff --git a/revision.h b/revision.h
index 89132df..5658ddd 100644
--- a/revision.h
+++ b/revision.h
@@ -169,6 +169,8 @@ struct rev_info {
unsigned long min_age;
int min_parents;
int max_parents;
+ int (*include_check)(struct commit *, void *);
+ void *include_check_data;
/* diff info for patches and for paths limiting */
struct diff_options diffopt;
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 06/21] sha1_file: export `git_open_noatime`
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (4 preceding siblings ...)
2013-11-14 12:43 ` [PATCH v3 05/21] revision: allow setting custom limiter function Jeff King
@ 2013-11-14 12:43 ` Jeff King
2013-11-14 12:43 ` [PATCH v3 07/21] compat: add endianness helpers Jeff King
` (15 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:43 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
The `git_open_noatime` helper can be of general interest for other
consumers of git's different on-disk formats.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
cache.h | 1 +
sha1_file.c | 4 +---
2 files changed, 2 insertions(+), 3 deletions(-)
diff --git a/cache.h b/cache.h
index ce377e1..6d4ef65 100644
--- a/cache.h
+++ b/cache.h
@@ -780,6 +780,7 @@ extern int hash_sha1_file(const void *buf, unsigned long len, const char *type,
extern int write_sha1_file(const void *buf, unsigned long len, const char *type, unsigned char *return_sha1);
extern int pretend_sha1_file(void *, unsigned long, enum object_type, unsigned char *);
extern int force_object_loose(const unsigned char *sha1, time_t mtime);
+extern int git_open_noatime(const char *name);
extern void *map_sha1_file(const unsigned char *sha1, unsigned long *size);
extern int unpack_sha1_header(git_zstream *stream, unsigned char *map, unsigned long mapsize, void *buffer, unsigned long bufsiz);
extern int parse_sha1_header(const char *hdr, unsigned long *sizep);
diff --git a/sha1_file.c b/sha1_file.c
index 7dadd04..5557bd9 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -239,8 +239,6 @@ char *sha1_pack_index_name(const unsigned char *sha1)
struct alternate_object_database *alt_odb_list;
static struct alternate_object_database **alt_odb_tail;
-static int git_open_noatime(const char *name);
-
/*
* Prepare alternate object database registry.
*
@@ -1357,7 +1355,7 @@ int check_sha1_signature(const unsigned char *sha1, void *map,
return hashcmp(sha1, real_sha1) ? -1 : 0;
}
-static int git_open_noatime(const char *name)
+int git_open_noatime(const char *name)
{
static int sha1_file_open_flag = O_NOATIME;
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 07/21] compat: add endianness helpers
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (5 preceding siblings ...)
2013-11-14 12:43 ` [PATCH v3 06/21] sha1_file: export `git_open_noatime` Jeff King
@ 2013-11-14 12:43 ` Jeff King
2013-11-14 12:43 ` [PATCH v3 08/21] ewah: compressed bitmap implementation Jeff King
` (14 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:43 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
The POSIX standard doesn't currently define a `ntohll`/`htonll`
function pair to perform network-to-host and host-to-network
swaps of 64-bit data. These 64-bit swaps are necessary for the on-disk
storage of EWAH bitmaps if they are not in native byte order.
Many thanks to Ramsay Jones <ramsay@ramsay1.demon.co.uk> and
Torsten Bögershausen <tboegi@web.de> for cygwin/mingw/msvc
portability fixes.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
compat/bswap.h | 76 +++++++++++++++++++++++++++++++++++++++++++++++++++++++++-
1 file changed, 75 insertions(+), 1 deletion(-)
diff --git a/compat/bswap.h b/compat/bswap.h
index 5061214..c18a78e 100644
--- a/compat/bswap.h
+++ b/compat/bswap.h
@@ -17,7 +17,20 @@ static inline uint32_t default_swab32(uint32_t val)
((val & 0x000000ff) << 24));
}
+static inline uint64_t default_bswap64(uint64_t val)
+{
+ return (((val & (uint64_t)0x00000000000000ffULL) << 56) |
+ ((val & (uint64_t)0x000000000000ff00ULL) << 40) |
+ ((val & (uint64_t)0x0000000000ff0000ULL) << 24) |
+ ((val & (uint64_t)0x00000000ff000000ULL) << 8) |
+ ((val & (uint64_t)0x000000ff00000000ULL) >> 8) |
+ ((val & (uint64_t)0x0000ff0000000000ULL) >> 24) |
+ ((val & (uint64_t)0x00ff000000000000ULL) >> 40) |
+ ((val & (uint64_t)0xff00000000000000ULL) >> 56));
+}
+
#undef bswap32
+#undef bswap64
#if defined(__GNUC__) && (defined(__i386__) || defined(__x86_64__))
@@ -32,15 +45,42 @@ static inline uint32_t git_bswap32(uint32_t x)
return result;
}
+#define bswap64 git_bswap64
+#if defined(__x86_64__)
+static inline uint64_t git_bswap64(uint64_t x)
+{
+ uint64_t result;
+ if (__builtin_constant_p(x))
+ result = default_bswap64(x);
+ else
+ __asm__("bswap %q0" : "=r" (result) : "0" (x));
+ return result;
+}
+#else
+static inline uint64_t git_bswap64(uint64_t x)
+{
+ union { uint64_t i64; uint32_t i32[2]; } tmp, result;
+ if (__builtin_constant_p(x))
+ result.i64 = default_bswap64(x);
+ else {
+ tmp.i64 = x;
+ result.i32[0] = git_bswap32(tmp.i32[1]);
+ result.i32[1] = git_bswap32(tmp.i32[0]);
+ }
+ return result.i64;
+}
+#endif
+
#elif defined(_MSC_VER) && (defined(_M_IX86) || defined(_M_X64))
#include <stdlib.h>
#define bswap32(x) _byteswap_ulong(x)
+#define bswap64(x) _byteswap_uint64(x)
#endif
-#ifdef bswap32
+#if defined(bswap32)
#undef ntohl
#undef htonl
@@ -48,3 +88,37 @@ static inline uint32_t git_bswap32(uint32_t x)
#define htonl(x) bswap32(x)
#endif
+
+#if defined(bswap64)
+
+#undef ntohll
+#undef htonll
+#define ntohll(x) bswap64(x)
+#define htonll(x) bswap64(x)
+
+#else
+
+#undef ntohll
+#undef htonll
+
+#if !defined(__BYTE_ORDER)
+# if defined(BYTE_ORDER) && defined(LITTLE_ENDIAN) && defined(BIG_ENDIAN)
+# define __BYTE_ORDER BYTE_ORDER
+# define __LITTLE_ENDIAN LITTLE_ENDIAN
+# define __BIG_ENDIAN BIG_ENDIAN
+# endif
+#endif
+
+#if !defined(__BYTE_ORDER)
+# error "Cannot determine endianness"
+#endif
+
+#if __BYTE_ORDER == __BIG_ENDIAN
+# define ntohll(n) (n)
+# define htonll(n) (n)
+#else
+# define ntohll(n) default_bswap64(n)
+# define htonll(n) default_bswap64(n)
+#endif
+
+#endif
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 08/21] ewah: compressed bitmap implementation
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (6 preceding siblings ...)
2013-11-14 12:43 ` [PATCH v3 07/21] compat: add endianness helpers Jeff King
@ 2013-11-14 12:43 ` Jeff King
2013-11-14 12:44 ` [PATCH v3 09/21] documentation: add documentation for the bitmap format Jeff King
` (13 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:43 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
EWAH is a word-aligned compressed variant of a bitset (i.e. a data
structure that acts as a 0-indexed boolean array for many entries).
It uses a 64-bit run-length encoding (RLE) compression scheme,
trading some compression for better processing speed.
The goal of this word-aligned implementation is not to achieve
the best compression, but rather to improve query processing time.
As it stands right now, this EWAH implementation will always be more
efficient storage-wise than its uncompressed alternative.
EWAH arrays will be used as the on-disk format to store reachability
bitmaps for all objects in a repository while keeping reasonable sizes,
in the same way that JGit does.
This EWAH implementation is a mostly straightforward port of the
original `javaewah` library that JGit currently uses. The library is
self-contained and has been embedded whole (4 files) inside the `ewah`
folder to ease redistribution.
The library is re-licensed under the GPLv2 with the permission of Daniel
Lemire, the original author. The source code for the C version can
be found on GitHub:
https://github.com/vmg/libewok
The original Java implementation can also be found on GitHub:
https://github.com/lemire/javaewah
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Makefile | 11 +-
ewah/bitmap.c | 221 ++++++++++++++++
ewah/ewah_bitmap.c | 726 +++++++++++++++++++++++++++++++++++++++++++++++++++++
ewah/ewah_io.c | 193 ++++++++++++++
ewah/ewah_rlw.c | 115 +++++++++
ewah/ewok.h | 235 +++++++++++++++++
ewah/ewok_rlw.h | 114 +++++++++
7 files changed, 1613 insertions(+), 2 deletions(-)
create mode 100644 ewah/bitmap.c
create mode 100644 ewah/ewah_bitmap.c
create mode 100644 ewah/ewah_io.c
create mode 100644 ewah/ewah_rlw.c
create mode 100644 ewah/ewok.h
create mode 100644 ewah/ewok_rlw.h
diff --git a/Makefile b/Makefile
index 48ff0bd..64a1ed7 100644
--- a/Makefile
+++ b/Makefile
@@ -667,6 +667,8 @@ LIB_H += diff.h
LIB_H += diffcore.h
LIB_H += dir.h
LIB_H += exec_cmd.h
+LIB_H += ewah/ewok.h
+LIB_H += ewah/ewok_rlw.h
LIB_H += fetch-pack.h
LIB_H += fmt-merge-msg.h
LIB_H += fsck.h
@@ -800,6 +802,10 @@ LIB_OBJS += dir.o
LIB_OBJS += editor.o
LIB_OBJS += entry.o
LIB_OBJS += environment.o
+LIB_OBJS += ewah/bitmap.o
+LIB_OBJS += ewah/ewah_bitmap.o
+LIB_OBJS += ewah/ewah_io.o
+LIB_OBJS += ewah/ewah_rlw.o
LIB_OBJS += exec_cmd.o
LIB_OBJS += fetch-pack.o
LIB_OBJS += fsck.o
@@ -2474,8 +2480,9 @@ profile-clean:
$(RM) $(addsuffix *.gcno,$(addprefix $(PROFILE_DIR)/, $(object_dirs)))
clean: profile-clean coverage-clean
- $(RM) *.o *.res block-sha1/*.o ppc/*.o compat/*.o compat/*/*.o xdiff/*.o vcs-svn/*.o \
- builtin/*.o $(LIB_FILE) $(XDIFF_LIB) $(VCSSVN_LIB)
+ $(RM) *.o *.res block-sha1/*.o ppc/*.o compat/*.o compat/*/*.o
+ $(RM) xdiff/*.o vcs-svn/*.o ewah/*.o builtin/*.o
+ $(RM) $(LIB_FILE) $(XDIFF_LIB) $(VCSSVN_LIB)
$(RM) $(ALL_PROGRAMS) $(SCRIPT_LIB) $(BUILT_INS) git$X
$(RM) $(TEST_PROGRAMS) $(NO_INSTALL)
$(RM) -r bin-wrappers $(dep_dirs)
diff --git a/ewah/bitmap.c b/ewah/bitmap.c
new file mode 100644
index 0000000..710e58c
--- /dev/null
+++ b/ewah/bitmap.c
@@ -0,0 +1,221 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#include "git-compat-util.h"
+#include "ewok.h"
+
+#define MASK(x) ((eword_t)1 << (x % BITS_IN_WORD))
+#define BLOCK(x) (x / BITS_IN_WORD)
+
+struct bitmap *bitmap_new(void)
+{
+ struct bitmap *bitmap = ewah_malloc(sizeof(struct bitmap));
+ bitmap->words = ewah_calloc(32, sizeof(eword_t));
+ bitmap->word_alloc = 32;
+ return bitmap;
+}
+
+void bitmap_set(struct bitmap *self, size_t pos)
+{
+ size_t block = BLOCK(pos);
+
+ if (block >= self->word_alloc) {
+ size_t old_size = self->word_alloc;
+ self->word_alloc = block * 2;
+ self->words = ewah_realloc(self->words,
+ self->word_alloc * sizeof(eword_t));
+
+ memset(self->words + old_size, 0x0,
+ (self->word_alloc - old_size) * sizeof(eword_t));
+ }
+
+ self->words[block] |= MASK(pos);
+}
+
+void bitmap_clear(struct bitmap *self, size_t pos)
+{
+ size_t block = BLOCK(pos);
+
+ if (block < self->word_alloc)
+ self->words[block] &= ~MASK(pos);
+}
+
+int bitmap_get(struct bitmap *self, size_t pos)
+{
+ size_t block = BLOCK(pos);
+ return block < self->word_alloc &&
+ (self->words[block] & MASK(pos)) != 0;
+}
+
+struct ewah_bitmap *bitmap_to_ewah(struct bitmap *bitmap)
+{
+ struct ewah_bitmap *ewah = ewah_new();
+ size_t i, running_empty_words = 0;
+ eword_t last_word = 0;
+
+ for (i = 0; i < bitmap->word_alloc; ++i) {
+ if (bitmap->words[i] == 0) {
+ running_empty_words++;
+ continue;
+ }
+
+ if (last_word != 0)
+ ewah_add(ewah, last_word);
+
+ if (running_empty_words > 0) {
+ ewah_add_empty_words(ewah, 0, running_empty_words);
+ running_empty_words = 0;
+ }
+
+ last_word = bitmap->words[i];
+ }
+
+ ewah_add(ewah, last_word);
+ return ewah;
+}
+
+struct bitmap *ewah_to_bitmap(struct ewah_bitmap *ewah)
+{
+ struct bitmap *bitmap = bitmap_new();
+ struct ewah_iterator it;
+ eword_t blowup;
+ size_t i = 0;
+
+ ewah_iterator_init(&it, ewah);
+
+ while (ewah_iterator_next(&blowup, &it)) {
+ if (i >= bitmap->word_alloc) {
+ bitmap->word_alloc *= 1.5;
+ bitmap->words = ewah_realloc(
+ bitmap->words, bitmap->word_alloc * sizeof(eword_t));
+ }
+
+ bitmap->words[i++] = blowup;
+ }
+
+ bitmap->word_alloc = i;
+ return bitmap;
+}
+
+void bitmap_and_not(struct bitmap *self, struct bitmap *other)
+{
+ const size_t count = (self->word_alloc < other->word_alloc) ?
+ self->word_alloc : other->word_alloc;
+
+ size_t i;
+
+ for (i = 0; i < count; ++i)
+ self->words[i] &= ~other->words[i];
+}
+
+void bitmap_or_ewah(struct bitmap *self, struct ewah_bitmap *other)
+{
+ size_t original_size = self->word_alloc;
+ size_t other_final = (other->bit_size / BITS_IN_WORD) + 1;
+ size_t i = 0;
+ struct ewah_iterator it;
+ eword_t word;
+
+ if (self->word_alloc < other_final) {
+ self->word_alloc = other_final;
+ self->words = ewah_realloc(self->words,
+ self->word_alloc * sizeof(eword_t));
+ memset(self->words + original_size, 0x0,
+ (self->word_alloc - original_size) * sizeof(eword_t));
+ }
+
+ ewah_iterator_init(&it, other);
+
+ while (ewah_iterator_next(&word, &it))
+ self->words[i++] |= word;
+}
+
+void bitmap_each_bit(struct bitmap *self, ewah_callback callback, void *data)
+{
+ size_t pos = 0, i;
+
+ for (i = 0; i < self->word_alloc; ++i) {
+ eword_t word = self->words[i];
+ uint32_t offset;
+
+ if (word == (eword_t)~0) {
+ for (offset = 0; offset < BITS_IN_WORD; ++offset)
+ callback(pos++, data);
+ } else {
+ for (offset = 0; offset < BITS_IN_WORD; ++offset) {
+ if ((word >> offset) == 0)
+ break;
+
+ offset += ewah_bit_ctz64(word >> offset);
+ callback(pos + offset, data);
+ }
+ pos += BITS_IN_WORD;
+ }
+ }
+}
+
+size_t bitmap_popcount(struct bitmap *self)
+{
+ size_t i, count = 0;
+
+ for (i = 0; i < self->word_alloc; ++i)
+ count += ewah_bit_popcount64(self->words[i]);
+
+ return count;
+}
+
+int bitmap_equals(struct bitmap *self, struct bitmap *other)
+{
+ struct bitmap *big, *small;
+ size_t i;
+
+ if (self->word_alloc < other->word_alloc) {
+ small = self;
+ big = other;
+ } else {
+ small = other;
+ big = self;
+ }
+
+ for (i = 0; i < small->word_alloc; ++i) {
+ if (small->words[i] != big->words[i])
+ return 0;
+ }
+
+ for (; i < big->word_alloc; ++i) {
+ if (big->words[i] != 0)
+ return 0;
+ }
+
+ return 1;
+}
+
+void bitmap_reset(struct bitmap *bitmap)
+{
+ memset(bitmap->words, 0x0, bitmap->word_alloc * sizeof(eword_t));
+}
+
+void bitmap_free(struct bitmap *bitmap)
+{
+ if (bitmap == NULL)
+ return;
+
+ free(bitmap->words);
+ free(bitmap);
+}
diff --git a/ewah/ewah_bitmap.c b/ewah/ewah_bitmap.c
new file mode 100644
index 0000000..f104b87
--- /dev/null
+++ b/ewah/ewah_bitmap.c
@@ -0,0 +1,726 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#include "git-compat-util.h"
+#include "ewok.h"
+#include "ewok_rlw.h"
+
+static inline size_t min_size(size_t a, size_t b)
+{
+ return a < b ? a : b;
+}
+
+static inline size_t max_size(size_t a, size_t b)
+{
+ return a > b ? a : b;
+}
+
+static inline void buffer_grow(struct ewah_bitmap *self, size_t new_size)
+{
+ size_t rlw_offset = (uint8_t *)self->rlw - (uint8_t *)self->buffer;
+
+ if (self->alloc_size >= new_size)
+ return;
+
+ self->alloc_size = new_size;
+ self->buffer = ewah_realloc(self->buffer,
+ self->alloc_size * sizeof(eword_t));
+ self->rlw = self->buffer + (rlw_offset / sizeof(size_t));
+}
+
+static inline void buffer_push(struct ewah_bitmap *self, eword_t value)
+{
+ if (self->buffer_size + 1 >= self->alloc_size)
+ buffer_grow(self, self->buffer_size * 3 / 2);
+
+ self->buffer[self->buffer_size++] = value;
+}
+
+static void buffer_push_rlw(struct ewah_bitmap *self, eword_t value)
+{
+ buffer_push(self, value);
+ self->rlw = self->buffer + self->buffer_size - 1;
+}
+
+static size_t add_empty_words(struct ewah_bitmap *self, int v, size_t number)
+{
+ size_t added = 0;
+ eword_t runlen, can_add;
+
+ if (rlw_get_run_bit(self->rlw) != v && rlw_size(self->rlw) == 0) {
+ rlw_set_run_bit(self->rlw, v);
+ } else if (rlw_get_literal_words(self->rlw) != 0 ||
+ rlw_get_run_bit(self->rlw) != v) {
+ buffer_push_rlw(self, 0);
+ if (v) rlw_set_run_bit(self->rlw, v);
+ added++;
+ }
+
+ runlen = rlw_get_running_len(self->rlw);
+ can_add = min_size(number, RLW_LARGEST_RUNNING_COUNT - runlen);
+
+ rlw_set_running_len(self->rlw, runlen + can_add);
+ number -= can_add;
+
+ while (number >= RLW_LARGEST_RUNNING_COUNT) {
+ buffer_push_rlw(self, 0);
+ added++;
+ if (v) rlw_set_run_bit(self->rlw, v);
+ rlw_set_running_len(self->rlw, RLW_LARGEST_RUNNING_COUNT);
+ number -= RLW_LARGEST_RUNNING_COUNT;
+ }
+
+ if (number > 0) {
+ buffer_push_rlw(self, 0);
+ added++;
+
+ if (v) rlw_set_run_bit(self->rlw, v);
+ rlw_set_running_len(self->rlw, number);
+ }
+
+ return added;
+}
+
+size_t ewah_add_empty_words(struct ewah_bitmap *self, int v, size_t number)
+{
+ if (number == 0)
+ return 0;
+
+ self->bit_size += number * BITS_IN_WORD;
+ return add_empty_words(self, v, number);
+}
+
+static size_t add_literal(struct ewah_bitmap *self, eword_t new_data)
+{
+ eword_t current_num = rlw_get_literal_words(self->rlw);
+
+ if (current_num >= RLW_LARGEST_LITERAL_COUNT) {
+ buffer_push_rlw(self, 0);
+
+ rlw_set_literal_words(self->rlw, 1);
+ buffer_push(self, new_data);
+ return 2;
+ }
+
+ rlw_set_literal_words(self->rlw, current_num + 1);
+
+ /* sanity check */
+ assert(rlw_get_literal_words(self->rlw) == current_num + 1);
+
+ buffer_push(self, new_data);
+ return 1;
+}
+
+void ewah_add_dirty_words(
+ struct ewah_bitmap *self, const eword_t *buffer,
+ size_t number, int negate)
+{
+ size_t literals, can_add;
+
+ while (1) {
+ literals = rlw_get_literal_words(self->rlw);
+ can_add = min_size(number, RLW_LARGEST_LITERAL_COUNT - literals);
+
+ rlw_set_literal_words(self->rlw, literals + can_add);
+
+ if (self->buffer_size + can_add >= self->alloc_size)
+ buffer_grow(self, (self->buffer_size + can_add) * 3 / 2);
+
+ if (negate) {
+ size_t i;
+ for (i = 0; i < can_add; ++i)
+ self->buffer[self->buffer_size++] = ~buffer[i];
+ } else {
+ memcpy(self->buffer + self->buffer_size,
+ buffer, can_add * sizeof(eword_t));
+ self->buffer_size += can_add;
+ }
+
+ self->bit_size += can_add * BITS_IN_WORD;
+
+ if (number - can_add == 0)
+ break;
+
+ buffer_push_rlw(self, 0);
+ buffer += can_add;
+ number -= can_add;
+ }
+}
+
+static size_t add_empty_word(struct ewah_bitmap *self, int v)
+{
+ int no_literal = (rlw_get_literal_words(self->rlw) == 0);
+ eword_t run_len = rlw_get_running_len(self->rlw);
+
+ if (no_literal && run_len == 0) {
+ rlw_set_run_bit(self->rlw, v);
+ assert(rlw_get_run_bit(self->rlw) == v);
+ }
+
+ if (no_literal && rlw_get_run_bit(self->rlw) == v &&
+ run_len < RLW_LARGEST_RUNNING_COUNT) {
+ rlw_set_running_len(self->rlw, run_len + 1);
+ assert(rlw_get_running_len(self->rlw) == run_len + 1);
+ return 0;
+ } else {
+ buffer_push_rlw(self, 0);
+
+ assert(rlw_get_running_len(self->rlw) == 0);
+ assert(rlw_get_run_bit(self->rlw) == 0);
+ assert(rlw_get_literal_words(self->rlw) == 0);
+
+ rlw_set_run_bit(self->rlw, v);
+ assert(rlw_get_run_bit(self->rlw) == v);
+
+ rlw_set_running_len(self->rlw, 1);
+ assert(rlw_get_running_len(self->rlw) == 1);
+ assert(rlw_get_literal_words(self->rlw) == 0);
+ return 1;
+ }
+}
+
+size_t ewah_add(struct ewah_bitmap *self, eword_t word)
+{
+ self->bit_size += BITS_IN_WORD;
+
+ if (word == 0)
+ return add_empty_word(self, 0);
+
+ if (word == (eword_t)(~0))
+ return add_empty_word(self, 1);
+
+ return add_literal(self, word);
+}
+
+void ewah_set(struct ewah_bitmap *self, size_t i)
+{
+ const size_t dist =
+ (i + BITS_IN_WORD) / BITS_IN_WORD -
+ (self->bit_size + BITS_IN_WORD - 1) / BITS_IN_WORD;
+
+ assert(i >= self->bit_size);
+
+ self->bit_size = i + 1;
+
+ if (dist > 0) {
+ if (dist > 1)
+ add_empty_words(self, 0, dist - 1);
+
+ add_literal(self, (eword_t)1 << (i % BITS_IN_WORD));
+ return;
+ }
+
+ if (rlw_get_literal_words(self->rlw) == 0) {
+ rlw_set_running_len(self->rlw,
+ rlw_get_running_len(self->rlw) - 1);
+ add_literal(self, (eword_t)1 << (i % BITS_IN_WORD));
+ return;
+ }
+
+ self->buffer[self->buffer_size - 1] |=
+ ((eword_t)1 << (i % BITS_IN_WORD));
+
+ /* check if we just completed a stream of 1s */
+ if (self->buffer[self->buffer_size - 1] == (eword_t)(~0)) {
+ self->buffer[--self->buffer_size] = 0;
+ rlw_set_literal_words(self->rlw,
+ rlw_get_literal_words(self->rlw) - 1);
+ add_empty_word(self, 1);
+ }
+}
+
+void ewah_each_bit(struct ewah_bitmap *self, void (*callback)(size_t, void*), void *payload)
+{
+ size_t pos = 0;
+ size_t pointer = 0;
+ size_t k;
+
+ while (pointer < self->buffer_size) {
+ eword_t *word = &self->buffer[pointer];
+
+ if (rlw_get_run_bit(word)) {
+ size_t len = rlw_get_running_len(word) * BITS_IN_WORD;
+ for (k = 0; k < len; ++k, ++pos)
+ callback(pos, payload);
+ } else {
+ pos += rlw_get_running_len(word) * BITS_IN_WORD;
+ }
+
+ ++pointer;
+
+ for (k = 0; k < rlw_get_literal_words(word); ++k) {
+ int c;
+
+ /* todo: zero count optimization */
+ for (c = 0; c < BITS_IN_WORD; ++c, ++pos) {
+ if ((self->buffer[pointer] & ((eword_t)1 << c)) != 0)
+ callback(pos, payload);
+ }
+
+ ++pointer;
+ }
+ }
+}
+
+struct ewah_bitmap *ewah_new(void)
+{
+ struct ewah_bitmap *self;
+
+ self = ewah_malloc(sizeof(struct ewah_bitmap));
+ if (self == NULL)
+ return NULL;
+
+ self->buffer = ewah_malloc(32 * sizeof(eword_t));
+ self->alloc_size = 32;
+
+ ewah_clear(self);
+ return self;
+}
+
+void ewah_clear(struct ewah_bitmap *self)
+{
+ self->buffer_size = 1;
+ self->buffer[0] = 0;
+ self->bit_size = 0;
+ self->rlw = self->buffer;
+}
+
+void ewah_free(struct ewah_bitmap *self)
+{
+ if (!self)
+ return;
+
+ if (self->alloc_size)
+ free(self->buffer);
+
+ free(self);
+}
+
+static void read_new_rlw(struct ewah_iterator *it)
+{
+ const eword_t *word = NULL;
+
+ it->literals = 0;
+ it->compressed = 0;
+
+ while (1) {
+ word = &it->buffer[it->pointer];
+
+ it->rl = rlw_get_running_len(word);
+ it->lw = rlw_get_literal_words(word);
+ it->b = rlw_get_run_bit(word);
+
+ if (it->rl || it->lw)
+ return;
+
+ if (it->pointer < it->buffer_size - 1) {
+ it->pointer++;
+ } else {
+ it->pointer = it->buffer_size;
+ return;
+ }
+ }
+}
+
+int ewah_iterator_next(eword_t *next, struct ewah_iterator *it)
+{
+ if (it->pointer >= it->buffer_size)
+ return 0;
+
+ if (it->compressed < it->rl) {
+ it->compressed++;
+ *next = it->b ? (eword_t)(~0) : 0;
+ } else {
+ assert(it->literals < it->lw);
+
+ it->literals++;
+ it->pointer++;
+
+ assert(it->pointer < it->buffer_size);
+
+ *next = it->buffer[it->pointer];
+ }
+
+ if (it->compressed == it->rl && it->literals == it->lw) {
+ if (++it->pointer < it->buffer_size)
+ read_new_rlw(it);
+ }
+
+ return 1;
+}
+
+void ewah_iterator_init(struct ewah_iterator *it, struct ewah_bitmap *parent)
+{
+ it->buffer = parent->buffer;
+ it->buffer_size = parent->buffer_size;
+ it->pointer = 0;
+
+ it->lw = 0;
+ it->rl = 0;
+ it->compressed = 0;
+ it->literals = 0;
+ it->b = 0;
+
+ if (it->pointer < it->buffer_size)
+ read_new_rlw(it);
+}
+
+void ewah_dump(struct ewah_bitmap *self)
+{
+ size_t i;
+ fprintf(stderr, "%"PRIuMAX" bits | %"PRIuMAX" words | ",
+ (uintmax_t)self->bit_size, (uintmax_t)self->buffer_size);
+
+ for (i = 0; i < self->buffer_size; ++i)
+ fprintf(stderr, "%016"PRIx64" ", (uint64_t)self->buffer[i]);
+
+ fprintf(stderr, "\n");
+}
+
+void ewah_not(struct ewah_bitmap *self)
+{
+ size_t pointer = 0;
+
+ while (pointer < self->buffer_size) {
+ eword_t *word = &self->buffer[pointer];
+ size_t literals, k;
+
+ rlw_xor_run_bit(word);
+ ++pointer;
+
+ literals = rlw_get_literal_words(word);
+ for (k = 0; k < literals; ++k) {
+ self->buffer[pointer] = ~self->buffer[pointer];
+ ++pointer;
+ }
+ }
+}
+
+void ewah_xor(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out)
+{
+ struct rlw_iterator rlw_i;
+ struct rlw_iterator rlw_j;
+ size_t literals;
+
+ rlwit_init(&rlw_i, ewah_i);
+ rlwit_init(&rlw_j, ewah_j);
+
+ while (rlwit_word_size(&rlw_i) > 0 && rlwit_word_size(&rlw_j) > 0) {
+ while (rlw_i.rlw.running_len > 0 || rlw_j.rlw.running_len > 0) {
+ struct rlw_iterator *prey, *predator;
+ size_t index;
+ int negate_words;
+
+ if (rlw_i.rlw.running_len < rlw_j.rlw.running_len) {
+ prey = &rlw_i;
+ predator = &rlw_j;
+ } else {
+ prey = &rlw_j;
+ predator = &rlw_i;
+ }
+
+ negate_words = !!predator->rlw.running_bit;
+ index = rlwit_discharge(prey, out,
+ predator->rlw.running_len, negate_words);
+
+ ewah_add_empty_words(out, negate_words,
+ predator->rlw.running_len - index);
+
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ }
+
+ literals = min_size(
+ rlw_i.rlw.literal_words,
+ rlw_j.rlw.literal_words);
+
+ if (literals) {
+ size_t k;
+
+ for (k = 0; k < literals; ++k) {
+ ewah_add(out,
+ rlw_i.buffer[rlw_i.literal_word_start + k] ^
+ rlw_j.buffer[rlw_j.literal_word_start + k]
+ );
+ }
+
+ rlwit_discard_first_words(&rlw_i, literals);
+ rlwit_discard_first_words(&rlw_j, literals);
+ }
+ }
+
+ if (rlwit_word_size(&rlw_i) > 0)
+ rlwit_discharge(&rlw_i, out, ~0, 0);
+ else
+ rlwit_discharge(&rlw_j, out, ~0, 0);
+
+ out->bit_size = max_size(ewah_i->bit_size, ewah_j->bit_size);
+}
+
+void ewah_and(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out)
+{
+ struct rlw_iterator rlw_i;
+ struct rlw_iterator rlw_j;
+ size_t literals;
+
+ rlwit_init(&rlw_i, ewah_i);
+ rlwit_init(&rlw_j, ewah_j);
+
+ while (rlwit_word_size(&rlw_i) > 0 && rlwit_word_size(&rlw_j) > 0) {
+ while (rlw_i.rlw.running_len > 0 || rlw_j.rlw.running_len > 0) {
+ struct rlw_iterator *prey, *predator;
+
+ if (rlw_i.rlw.running_len < rlw_j.rlw.running_len) {
+ prey = &rlw_i;
+ predator = &rlw_j;
+ } else {
+ prey = &rlw_j;
+ predator = &rlw_i;
+ }
+
+ if (predator->rlw.running_bit == 0) {
+ ewah_add_empty_words(out, 0,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(prey,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ } else {
+ size_t index = rlwit_discharge(prey, out,
+ predator->rlw.running_len, 0);
+ ewah_add_empty_words(out, 0,
+ predator->rlw.running_len - index);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ }
+ }
+
+ literals = min_size(
+ rlw_i.rlw.literal_words,
+ rlw_j.rlw.literal_words);
+
+ if (literals) {
+ size_t k;
+
+ for (k = 0; k < literals; ++k) {
+ ewah_add(out,
+ rlw_i.buffer[rlw_i.literal_word_start + k] &
+ rlw_j.buffer[rlw_j.literal_word_start + k]
+ );
+ }
+
+ rlwit_discard_first_words(&rlw_i, literals);
+ rlwit_discard_first_words(&rlw_j, literals);
+ }
+ }
+
+ if (rlwit_word_size(&rlw_i) > 0)
+ rlwit_discharge_empty(&rlw_i, out);
+ else
+ rlwit_discharge_empty(&rlw_j, out);
+
+ out->bit_size = max_size(ewah_i->bit_size, ewah_j->bit_size);
+}
+
+void ewah_and_not(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out)
+{
+ struct rlw_iterator rlw_i;
+ struct rlw_iterator rlw_j;
+ size_t literals;
+
+ rlwit_init(&rlw_i, ewah_i);
+ rlwit_init(&rlw_j, ewah_j);
+
+ while (rlwit_word_size(&rlw_i) > 0 && rlwit_word_size(&rlw_j) > 0) {
+ while (rlw_i.rlw.running_len > 0 || rlw_j.rlw.running_len > 0) {
+ struct rlw_iterator *prey, *predator;
+
+ if (rlw_i.rlw.running_len < rlw_j.rlw.running_len) {
+ prey = &rlw_i;
+ predator = &rlw_j;
+ } else {
+ prey = &rlw_j;
+ predator = &rlw_i;
+ }
+
+ if ((predator->rlw.running_bit && prey == &rlw_i) ||
+ (!predator->rlw.running_bit && prey != &rlw_i)) {
+ ewah_add_empty_words(out, 0,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(prey,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ } else {
+ size_t index;
+ int negate_words;
+
+ negate_words = (&rlw_i != prey);
+ index = rlwit_discharge(prey, out,
+ predator->rlw.running_len, negate_words);
+ ewah_add_empty_words(out, negate_words,
+ predator->rlw.running_len - index);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ }
+ }
+
+ literals = min_size(
+ rlw_i.rlw.literal_words,
+ rlw_j.rlw.literal_words);
+
+ if (literals) {
+ size_t k;
+
+ for (k = 0; k < literals; ++k) {
+ ewah_add(out,
+ rlw_i.buffer[rlw_i.literal_word_start + k] &
+ ~(rlw_j.buffer[rlw_j.literal_word_start + k])
+ );
+ }
+
+ rlwit_discard_first_words(&rlw_i, literals);
+ rlwit_discard_first_words(&rlw_j, literals);
+ }
+ }
+
+ if (rlwit_word_size(&rlw_i) > 0)
+ rlwit_discharge(&rlw_i, out, ~0, 0);
+ else
+ rlwit_discharge_empty(&rlw_j, out);
+
+ out->bit_size = max_size(ewah_i->bit_size, ewah_j->bit_size);
+}
+
+void ewah_or(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out)
+{
+ struct rlw_iterator rlw_i;
+ struct rlw_iterator rlw_j;
+ size_t literals;
+
+ rlwit_init(&rlw_i, ewah_i);
+ rlwit_init(&rlw_j, ewah_j);
+
+ while (rlwit_word_size(&rlw_i) > 0 && rlwit_word_size(&rlw_j) > 0) {
+ while (rlw_i.rlw.running_len > 0 || rlw_j.rlw.running_len > 0) {
+ struct rlw_iterator *prey, *predator;
+
+ if (rlw_i.rlw.running_len < rlw_j.rlw.running_len) {
+ prey = &rlw_i;
+ predator = &rlw_j;
+ } else {
+ prey = &rlw_j;
+ predator = &rlw_i;
+ }
+
+ if (predator->rlw.running_bit) {
+ ewah_add_empty_words(out, 0,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(prey,
+ predator->rlw.running_len);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ } else {
+ size_t index = rlwit_discharge(prey, out,
+ predator->rlw.running_len, 0);
+ ewah_add_empty_words(out, 0,
+ predator->rlw.running_len - index);
+ rlwit_discard_first_words(predator,
+ predator->rlw.running_len);
+ }
+ }
+
+ literals = min_size(
+ rlw_i.rlw.literal_words,
+ rlw_j.rlw.literal_words);
+
+ if (literals) {
+ size_t k;
+
+ for (k = 0; k < literals; ++k) {
+ ewah_add(out,
+ rlw_i.buffer[rlw_i.literal_word_start + k] |
+ rlw_j.buffer[rlw_j.literal_word_start + k]
+ );
+ }
+
+ rlwit_discard_first_words(&rlw_i, literals);
+ rlwit_discard_first_words(&rlw_j, literals);
+ }
+ }
+
+ if (rlwit_word_size(&rlw_i) > 0)
+ rlwit_discharge(&rlw_i, out, ~0, 0);
+ else
+ rlwit_discharge(&rlw_j, out, ~0, 0);
+
+ out->bit_size = max_size(ewah_i->bit_size, ewah_j->bit_size);
+}
+
+
+#define BITMAP_POOL_MAX 16
+static struct ewah_bitmap *bitmap_pool[BITMAP_POOL_MAX];
+static size_t bitmap_pool_size;
+
+struct ewah_bitmap *ewah_pool_new(void)
+{
+ if (bitmap_pool_size)
+ return bitmap_pool[--bitmap_pool_size];
+
+ return ewah_new();
+}
+
+void ewah_pool_free(struct ewah_bitmap *self)
+{
+ if (self == NULL)
+ return;
+
+ if (bitmap_pool_size == BITMAP_POOL_MAX ||
+ self->alloc_size == 0) {
+ ewah_free(self);
+ return;
+ }
+
+ ewah_clear(self);
+ bitmap_pool[bitmap_pool_size++] = self;
+}
+
+uint32_t ewah_checksum(struct ewah_bitmap *self)
+{
+ const uint8_t *p = (uint8_t *)self->buffer;
+ uint32_t crc = (uint32_t)self->bit_size;
+ size_t size = self->buffer_size * sizeof(eword_t);
+
+ while (size--)
+ crc = (crc << 5) - crc + (uint32_t)*p++;
+
+ return crc;
+}
diff --git a/ewah/ewah_io.c b/ewah/ewah_io.c
new file mode 100644
index 0000000..aed0da6
--- /dev/null
+++ b/ewah/ewah_io.c
@@ -0,0 +1,193 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#include "git-compat-util.h"
+#include "ewok.h"
+
+int ewah_serialize_native(struct ewah_bitmap *self, int fd)
+{
+ uint32_t write32;
+ size_t to_write = self->buffer_size * 8;
+
+ /* 32 bit -- bit size for the map */
+ write32 = (uint32_t)self->bit_size;
+ if (write(fd, &write32, 4) != 4)
+ return -1;
+
+ /** 32 bit -- number of compressed 64-bit words */
+ write32 = (uint32_t)self->buffer_size;
+ if (write(fd, &write32, 4) != 4)
+ return -1;
+
+ if (write(fd, self->buffer, to_write) != to_write)
+ return -1;
+
+ /** 32 bit -- position for the RLW */
+ write32 = self->rlw - self->buffer;
+ if (write(fd, &write32, 4) != 4)
+ return -1;
+
+ return (3 * 4) + to_write;
+}
+
+int ewah_serialize_to(struct ewah_bitmap *self,
+ int (*write_fun)(void *, const void *, size_t),
+ void *data)
+{
+ size_t i;
+ eword_t dump[2048];
+ const size_t words_per_dump = sizeof(dump) / sizeof(eword_t);
+ uint32_t bitsize, word_count, rlw_pos;
+
+ const eword_t *buffer;
+ size_t words_left;
+
+ /* 32 bit -- bit size for the map */
+ bitsize = htonl((uint32_t)self->bit_size);
+ if (write_fun(data, &bitsize, 4) != 4)
+ return -1;
+
+ /** 32 bit -- number of compressed 64-bit words */
+ word_count = htonl((uint32_t)self->buffer_size);
+ if (write_fun(data, &word_count, 4) != 4)
+ return -1;
+
+ /** 64 bit x N -- compressed words */
+ buffer = self->buffer;
+ words_left = self->buffer_size;
+
+ while (words_left >= words_per_dump) {
+ for (i = 0; i < words_per_dump; ++i, ++buffer)
+ dump[i] = htonll(*buffer);
+
+ if (write_fun(data, dump, sizeof(dump)) != sizeof(dump))
+ return -1;
+
+ words_left -= words_per_dump;
+ }
+
+ if (words_left) {
+ for (i = 0; i < words_left; ++i, ++buffer)
+ dump[i] = htonll(*buffer);
+
+ if (write_fun(data, dump, words_left * 8) != words_left * 8)
+ return -1;
+ }
+
+ /** 32 bit -- position for the RLW */
+ rlw_pos = (uint8_t*)self->rlw - (uint8_t *)self->buffer;
+ rlw_pos = htonl(rlw_pos / sizeof(eword_t));
+
+ if (write_fun(data, &rlw_pos, 4) != 4)
+ return -1;
+
+ return (3 * 4) + (self->buffer_size * 8);
+}
+
+static int write_helper(void *fd, const void *buf, size_t len)
+{
+ return write((intptr_t)fd, buf, len);
+}
+
+int ewah_serialize(struct ewah_bitmap *self, int fd)
+{
+ return ewah_serialize_to(self, write_helper, (void *)(intptr_t)fd);
+}
+
+int ewah_read_mmap(struct ewah_bitmap *self, void *map, size_t len)
+{
+ uint32_t *read32 = map;
+ eword_t *read64;
+ size_t i;
+
+ self->bit_size = ntohl(*read32++);
+ self->buffer_size = self->alloc_size = ntohl(*read32++);
+ self->buffer = ewah_realloc(self->buffer,
+ self->alloc_size * sizeof(eword_t));
+
+ if (!self->buffer)
+ return -1;
+
+ for (i = 0, read64 = (void *)read32; i < self->buffer_size; ++i)
+ self->buffer[i] = ntohll(*read64++);
+
+ read32 = (void *)read64;
+ self->rlw = self->buffer + ntohl(*read32++);
+
+ return (3 * 4) + (self->buffer_size * 8);
+}
+
+int ewah_deserialize(struct ewah_bitmap *self, int fd)
+{
+ size_t i;
+ eword_t dump[2048];
+ const size_t words_per_dump = sizeof(dump) / sizeof(eword_t);
+ uint32_t bitsize, word_count, rlw_pos;
+
+ eword_t *buffer = NULL;
+ size_t words_left;
+
+ ewah_clear(self);
+
+ /* 32 bit -- bit size for the map */
+ if (read(fd, &bitsize, 4) != 4)
+ return -1;
+
+ self->bit_size = (size_t)ntohl(bitsize);
+
+ /** 32 bit -- number of compressed 64-bit words */
+ if (read(fd, &word_count, 4) != 4)
+ return -1;
+
+ self->buffer_size = self->alloc_size = (size_t)ntohl(word_count);
+ self->buffer = ewah_realloc(self->buffer,
+ self->alloc_size * sizeof(eword_t));
+
+ if (!self->buffer)
+ return -1;
+
+ /** 64 bit x N -- compressed words */
+ buffer = self->buffer;
+ words_left = self->buffer_size;
+
+ while (words_left >= words_per_dump) {
+ if (read(fd, dump, sizeof(dump)) != sizeof(dump))
+ return -1;
+
+ for (i = 0; i < words_per_dump; ++i, ++buffer)
+ *buffer = ntohll(dump[i]);
+
+ words_left -= words_per_dump;
+ }
+
+ if (words_left) {
+ if (read(fd, dump, words_left * 8) != words_left * 8)
+ return -1;
+
+ for (i = 0; i < words_left; ++i, ++buffer)
+ *buffer = ntohll(dump[i]);
+ }
+
+ /** 32 bit -- position for the RLW */
+ if (read(fd, &rlw_pos, 4) != 4)
+ return -1;
+
+ self->rlw = self->buffer + ntohl(rlw_pos);
+ return 0;
+}
diff --git a/ewah/ewah_rlw.c b/ewah/ewah_rlw.c
new file mode 100644
index 0000000..c723f1a
--- /dev/null
+++ b/ewah/ewah_rlw.c
@@ -0,0 +1,115 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#include "git-compat-util.h"
+#include "ewok.h"
+#include "ewok_rlw.h"
+
+static inline int next_word(struct rlw_iterator *it)
+{
+ if (it->pointer >= it->size)
+ return 0;
+
+ it->rlw.word = &it->buffer[it->pointer];
+ it->pointer += rlw_get_literal_words(it->rlw.word) + 1;
+
+ it->rlw.literal_words = rlw_get_literal_words(it->rlw.word);
+ it->rlw.running_len = rlw_get_running_len(it->rlw.word);
+ it->rlw.running_bit = rlw_get_run_bit(it->rlw.word);
+ it->rlw.literal_word_offset = 0;
+
+ return 1;
+}
+
+void rlwit_init(struct rlw_iterator *it, struct ewah_bitmap *from_ewah)
+{
+ it->buffer = from_ewah->buffer;
+ it->size = from_ewah->buffer_size;
+ it->pointer = 0;
+
+ next_word(it);
+
+ it->literal_word_start = rlwit_literal_words(it) +
+ it->rlw.literal_word_offset;
+}
+
+void rlwit_discard_first_words(struct rlw_iterator *it, size_t x)
+{
+ while (x > 0) {
+ size_t discard;
+
+ if (it->rlw.running_len > x) {
+ it->rlw.running_len -= x;
+ return;
+ }
+
+ x -= it->rlw.running_len;
+ it->rlw.running_len = 0;
+
+ discard = (x > it->rlw.literal_words) ? it->rlw.literal_words : x;
+
+ it->literal_word_start += discard;
+ it->rlw.literal_words -= discard;
+ x -= discard;
+
+ if (x > 0 || rlwit_word_size(it) == 0) {
+ if (!next_word(it))
+ break;
+
+ it->literal_word_start =
+ rlwit_literal_words(it) + it->rlw.literal_word_offset;
+ }
+ }
+}
+
+size_t rlwit_discharge(
+ struct rlw_iterator *it, struct ewah_bitmap *out, size_t max, int negate)
+{
+ size_t index = 0;
+
+ while (index < max && rlwit_word_size(it) > 0) {
+ size_t pd, pl = it->rlw.running_len;
+
+ if (index + pl > max)
+ pl = max - index;
+
+ ewah_add_empty_words(out, it->rlw.running_bit ^ negate, pl);
+ index += pl;
+
+ pd = it->rlw.literal_words;
+ if (pd + index > max)
+ pd = max - index;
+
+ ewah_add_dirty_words(out,
+ it->buffer + it->literal_word_start, pd, negate);
+
+ rlwit_discard_first_words(it, pd + pl);
+ index += pd;
+ }
+
+ return index;
+}
+
+void rlwit_discharge_empty(struct rlw_iterator *it, struct ewah_bitmap *out)
+{
+ while (rlwit_word_size(it) > 0) {
+ ewah_add_empty_words(out, 0, rlwit_word_size(it));
+ rlwit_discard_first_words(it, rlwit_word_size(it));
+ }
+}
diff --git a/ewah/ewok.h b/ewah/ewok.h
new file mode 100644
index 0000000..619afaa
--- /dev/null
+++ b/ewah/ewok.h
@@ -0,0 +1,235 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#ifndef __EWOK_BITMAP_H__
+#define __EWOK_BITMAP_H__
+
+#ifndef ewah_malloc
+# define ewah_malloc xmalloc
+#endif
+#ifndef ewah_realloc
+# define ewah_realloc xrealloc
+#endif
+#ifndef ewah_calloc
+# define ewah_calloc xcalloc
+#endif
+
+typedef uint64_t eword_t;
+#define BITS_IN_WORD (sizeof(eword_t) * 8)
+
+/**
+ * Do not use __builtin_popcountll. The GCC implementation
+ * is notoriously slow on all platforms.
+ *
+ * See: http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36041
+ */
+static inline uint32_t ewah_bit_popcount64(uint64_t x)
+{
+ x = (x & 0x5555555555555555ULL) + ((x >> 1) & 0x5555555555555555ULL);
+ x = (x & 0x3333333333333333ULL) + ((x >> 2) & 0x3333333333333333ULL);
+ x = (x & 0x0F0F0F0F0F0F0F0FULL) + ((x >> 4) & 0x0F0F0F0F0F0F0F0FULL);
+ return (x * 0x0101010101010101ULL) >> 56;
+}
+
+#ifdef __GNUC__
+#define ewah_bit_ctz64(x) __builtin_ctzll(x)
+#else
+static inline int ewah_bit_ctz64(uint64_t x)
+{
+ int n = 0;
+ if ((x & 0xffffffff) == 0) { x >>= 32; n += 32; }
+ if ((x & 0xffff) == 0) { x >>= 16; n += 16; }
+ if ((x & 0xff) == 0) { x >>= 8; n += 8; }
+ if ((x & 0xf) == 0) { x >>= 4; n += 4; }
+ if ((x & 0x3) == 0) { x >>= 2; n += 2; }
+ if ((x & 0x1) == 0) { x >>= 1; n += 1; }
+ return n + !x;
+}
+#endif
+
+struct ewah_bitmap {
+ eword_t *buffer;
+ size_t buffer_size;
+ size_t alloc_size;
+ size_t bit_size;
+ eword_t *rlw;
+};
+
+typedef void (*ewah_callback)(size_t pos, void *);
+
+struct ewah_bitmap *ewah_pool_new(void);
+void ewah_pool_free(struct ewah_bitmap *self);
+
+/**
+ * Allocate a new EWAH Compressed bitmap
+ */
+struct ewah_bitmap *ewah_new(void);
+
+/**
+ * Clear all the bits in the bitmap. Does not free or resize
+ * memory.
+ */
+void ewah_clear(struct ewah_bitmap *self);
+
+/**
+ * Free all the memory of the bitmap
+ */
+void ewah_free(struct ewah_bitmap *self);
+
+int ewah_serialize_to(struct ewah_bitmap *self,
+ int (*write_fun)(void *out, const void *buf, size_t len),
+ void *out);
+int ewah_serialize(struct ewah_bitmap *self, int fd);
+int ewah_serialize_native(struct ewah_bitmap *self, int fd);
+
+int ewah_deserialize(struct ewah_bitmap *self, int fd);
+int ewah_read_mmap(struct ewah_bitmap *self, void *map, size_t len);
+int ewah_read_mmap_native(struct ewah_bitmap *self, void *map, size_t len);
+
+uint32_t ewah_checksum(struct ewah_bitmap *self);
+
+/**
+ * Logical not (bitwise negation) in-place on the bitmap
+ *
+ * This operation is linear time based on the size of the bitmap.
+ */
+void ewah_not(struct ewah_bitmap *self);
+
+/**
+ * Call the given callback with the position of every single bit
+ * that has been set on the bitmap.
+ *
+ * This is an efficient operation that does not fully decompress
+ * the bitmap.
+ */
+void ewah_each_bit(struct ewah_bitmap *self, ewah_callback callback, void *payload);
+
+/**
+ * Set a given bit on the bitmap.
+ *
+ * The bit at position `pos` will be set to true. Because of the
+ * way that the bitmap is compressed, a set bit cannot be unset
+ * later on.
+ *
+ * Furthermore, since the bitmap uses streaming compression, bits
+ * can only set incrementally.
+ *
+ * E.g.
+ * ewah_set(bitmap, 1); // ok
+ * ewah_set(bitmap, 76); // ok
+ * ewah_set(bitmap, 77); // ok
+ * ewah_set(bitmap, 8712800127); // ok
+ * ewah_set(bitmap, 25); // failed, assert raised
+ */
+void ewah_set(struct ewah_bitmap *self, size_t i);
+
+struct ewah_iterator {
+ const eword_t *buffer;
+ size_t buffer_size;
+
+ size_t pointer;
+ eword_t compressed, literals;
+ eword_t rl, lw;
+ int b;
+};
+
+/**
+ * Initialize a new iterator to run through the bitmap in uncompressed form.
+ *
+ * The iterator can be stack allocated. The underlying bitmap must not be freed
+ * before the iteration is over.
+ *
+ * E.g.
+ *
+ * struct ewah_bitmap *bitmap = ewah_new();
+ * struct ewah_iterator it;
+ *
+ * ewah_iterator_init(&it, bitmap);
+ */
+void ewah_iterator_init(struct ewah_iterator *it, struct ewah_bitmap *parent);
+
+/**
+ * Yield every single word in the bitmap in uncompressed form. This is:
+ * yield single words (32-64 bits) where each bit represents an actual
+ * bit from the bitmap.
+ *
+ * Return: true if a word was yield, false if there are no words left
+ */
+int ewah_iterator_next(eword_t *next, struct ewah_iterator *it);
+
+void ewah_or(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out);
+
+void ewah_and_not(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out);
+
+void ewah_xor(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out);
+
+void ewah_and(
+ struct ewah_bitmap *ewah_i,
+ struct ewah_bitmap *ewah_j,
+ struct ewah_bitmap *out);
+
+void ewah_dump(struct ewah_bitmap *self);
+
+/**
+ * Direct word access
+ */
+size_t ewah_add_empty_words(struct ewah_bitmap *self, int v, size_t number);
+void ewah_add_dirty_words(
+ struct ewah_bitmap *self, const eword_t *buffer, size_t number, int negate);
+size_t ewah_add(struct ewah_bitmap *self, eword_t word);
+
+
+/**
+ * Uncompressed, old-school bitmap that can be efficiently compressed
+ * into an `ewah_bitmap`.
+ */
+struct bitmap {
+ eword_t *words;
+ size_t word_alloc;
+};
+
+struct bitmap *bitmap_new(void);
+void bitmap_set(struct bitmap *self, size_t pos);
+void bitmap_clear(struct bitmap *self, size_t pos);
+int bitmap_get(struct bitmap *self, size_t pos);
+void bitmap_reset(struct bitmap *self);
+void bitmap_free(struct bitmap *self);
+int bitmap_equals(struct bitmap *self, struct bitmap *other);
+int bitmap_is_subset(struct bitmap *self, struct bitmap *super);
+
+struct ewah_bitmap * bitmap_to_ewah(struct bitmap *bitmap);
+struct bitmap *ewah_to_bitmap(struct ewah_bitmap *ewah);
+
+void bitmap_and_not(struct bitmap *self, struct bitmap *other);
+void bitmap_or_ewah(struct bitmap *self, struct ewah_bitmap *other);
+void bitmap_or(struct bitmap *self, const struct bitmap *other);
+
+void bitmap_each_bit(struct bitmap *self, ewah_callback callback, void *data);
+size_t bitmap_popcount(struct bitmap *self);
+
+#endif
diff --git a/ewah/ewok_rlw.h b/ewah/ewok_rlw.h
new file mode 100644
index 0000000..63efdf9
--- /dev/null
+++ b/ewah/ewok_rlw.h
@@ -0,0 +1,114 @@
+/**
+ * Copyright 2013, GitHub, Inc
+ * Copyright 2009-2013, Daniel Lemire, Cliff Moon,
+ * David McIntosh, Robert Becho, Google Inc. and Veronika Zenz
+ *
+ * This program is free software; you can redistribute it and/or
+ * modify it under the terms of the GNU General Public License
+ * as published by the Free Software Foundation; either version 2
+ * of the License, or (at your option) any later version.
+ *
+ * This program is distributed in the hope that it will be useful,
+ * but WITHOUT ANY WARRANTY; without even the implied warranty of
+ * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
+ * GNU General Public License for more details.
+ *
+ * You should have received a copy of the GNU General Public License
+ * along with this program; if not, write to the Free Software
+ * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
+ */
+#ifndef __EWOK_RLW_H__
+#define __EWOK_RLW_H__
+
+#define RLW_RUNNING_BITS (sizeof(eword_t) * 4)
+#define RLW_LITERAL_BITS (sizeof(eword_t) * 8 - 1 - RLW_RUNNING_BITS)
+
+#define RLW_LARGEST_RUNNING_COUNT (((eword_t)1 << RLW_RUNNING_BITS) - 1)
+#define RLW_LARGEST_LITERAL_COUNT (((eword_t)1 << RLW_LITERAL_BITS) - 1)
+
+#define RLW_LARGEST_RUNNING_COUNT_SHIFT (RLW_LARGEST_RUNNING_COUNT << 1)
+
+#define RLW_RUNNING_LEN_PLUS_BIT (((eword_t)1 << (RLW_RUNNING_BITS + 1)) - 1)
+
+static int rlw_get_run_bit(const eword_t *word)
+{
+ return *word & (eword_t)1;
+}
+
+static inline void rlw_set_run_bit(eword_t *word, int b)
+{
+ if (b) {
+ *word |= (eword_t)1;
+ } else {
+ *word &= (eword_t)(~1);
+ }
+}
+
+static inline void rlw_xor_run_bit(eword_t *word)
+{
+ if (*word & 1) {
+ *word &= (eword_t)(~1);
+ } else {
+ *word |= (eword_t)1;
+ }
+}
+
+static inline void rlw_set_running_len(eword_t *word, eword_t l)
+{
+ *word |= RLW_LARGEST_RUNNING_COUNT_SHIFT;
+ *word &= (l << 1) | (~RLW_LARGEST_RUNNING_COUNT_SHIFT);
+}
+
+static inline eword_t rlw_get_running_len(const eword_t *word)
+{
+ return (*word >> 1) & RLW_LARGEST_RUNNING_COUNT;
+}
+
+static inline eword_t rlw_get_literal_words(const eword_t *word)
+{
+ return *word >> (1 + RLW_RUNNING_BITS);
+}
+
+static inline void rlw_set_literal_words(eword_t *word, eword_t l)
+{
+ *word |= ~RLW_RUNNING_LEN_PLUS_BIT;
+ *word &= (l << (RLW_RUNNING_BITS + 1)) | RLW_RUNNING_LEN_PLUS_BIT;
+}
+
+static inline eword_t rlw_size(const eword_t *self)
+{
+ return rlw_get_running_len(self) + rlw_get_literal_words(self);
+}
+
+struct rlw_iterator {
+ const eword_t *buffer;
+ size_t size;
+ size_t pointer;
+ size_t literal_word_start;
+
+ struct {
+ const eword_t *word;
+ int literal_words;
+ int running_len;
+ int literal_word_offset;
+ int running_bit;
+ } rlw;
+};
+
+void rlwit_init(struct rlw_iterator *it, struct ewah_bitmap *bitmap);
+void rlwit_discard_first_words(struct rlw_iterator *it, size_t x);
+size_t rlwit_discharge(
+ struct rlw_iterator *it, struct ewah_bitmap *out, size_t max, int negate);
+void rlwit_discharge_empty(struct rlw_iterator *it, struct ewah_bitmap *out);
+
+static inline size_t rlwit_word_size(struct rlw_iterator *it)
+{
+ return it->rlw.running_len + it->rlw.literal_words;
+}
+
+static inline size_t rlwit_literal_words(struct rlw_iterator *it)
+{
+ return it->pointer - it->rlw.literal_words;
+}
+
+#endif
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 09/21] documentation: add documentation for the bitmap format
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (7 preceding siblings ...)
2013-11-14 12:43 ` [PATCH v3 08/21] ewah: compressed bitmap implementation Jeff King
@ 2013-11-14 12:44 ` Jeff King
2013-11-14 12:44 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Jeff King
` (12 subsequent siblings)
21 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:44 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
This is the technical documentation for the JGit-compatible Bitmap v1
on-disk format.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/technical/bitmap-format.txt | 131 ++++++++++++++++++++++++++++++
1 file changed, 131 insertions(+)
create mode 100644 Documentation/technical/bitmap-format.txt
diff --git a/Documentation/technical/bitmap-format.txt b/Documentation/technical/bitmap-format.txt
new file mode 100644
index 0000000..7a86bd7
--- /dev/null
+++ b/Documentation/technical/bitmap-format.txt
@@ -0,0 +1,131 @@
+GIT bitmap v1 format
+====================
+
+ - A header appears at the beginning:
+
+ 4-byte signature: {'B', 'I', 'T', 'M'}
+
+ 2-byte version number (network byte order)
+ The current implementation only supports version 1
+ of the bitmap index (the same one as JGit).
+
+ 2-byte flags (network byte order)
+
+ The following flags are supported:
+
+ - BITMAP_OPT_FULL_DAG (0x1) REQUIRED
+ This flag must always be present. It implies that the bitmap
+ index has been generated for a packfile with full closure
+ (i.e. where every single object in the packfile can find
+ its parent links inside the same packfile). This is a
+ requirement for the bitmap index format, also present in JGit,
+ that greatly reduces the complexity of the implementation.
+
+ 4-byte entry count (network byte order)
+
+ The total count of entries (bitmapped commits) in this bitmap index.
+
+ 20-byte checksum
+
+ The SHA1 checksum of the pack this bitmap index belongs to.
+
+ - 4 EWAH bitmaps that act as type indexes
+
+ Type indexes are serialized after the hash cache in the shape
+ of four EWAH bitmaps stored consecutively (see Appendix A for
+ the serialization format of an EWAH bitmap).
+
+ There is a bitmap for each Git object type, stored in the following
+ order:
+
+ - Commits
+ - Trees
+ - Blobs
+ - Tags
+
+ In each bitmap, the `n`th bit is set to true if the `n`th object
+ in the packfile is of that type.
+
+ The obvious consequence is that the OR of all 4 bitmaps will result
+ in a full set (all bits set), and the AND of all 4 bitmaps will
+ result in an empty bitmap (no bits set).
+
+ - N entries with compressed bitmaps, one for each indexed commit
+
+ Where `N` is the total amount of entries in this bitmap index.
+ Each entry contains the following:
+
+ - 4-byte object position (network byte order)
+ The position **in the index for the packfile** where the
+ bitmap for this commit is found.
+
+ - 1-byte XOR-offset
+ The xor offset used to compress this bitmap. For an entry
+ in position `x`, a XOR offset of `y` means that the actual
+ bitmap representing this commit is composed by XORing the
+ bitmap for this entry with the bitmap in entry `x-y` (i.e.
+ the bitmap `y` entries before this one).
+
+ Note that this compression can be recursive. In order to
+ XOR this entry with a previous one, the previous entry needs
+ to be decompressed first, and so on.
+
+ The hard-limit for this offset is 160 (an entry can only be
+ xor'ed against one of the 160 entries preceding it). This
+ number is always positive, and hence entries are always xor'ed
+ with **previous** bitmaps, not bitmaps that will come afterwards
+ in the index.
+
+ - 1-byte flags for this bitmap
+ At the moment the only available flag is `0x1`, which hints
+ that this bitmap can be re-used when rebuilding bitmap indexes
+ for the repository.
+
+ - The compressed bitmap itself, see Appendix A.
+
+== Appendix A: Serialization format for an EWAH bitmap
+
+Ewah bitmaps are serialized in the same protocol as the JAVAEWAH
+library, making them backwards compatible with the JGit
+implementation:
+
+ - 4-byte number of bits of the resulting UNCOMPRESSED bitmap
+
+ - 4-byte number of words of the COMPRESSED bitmap, when stored
+
+ - N x 8-byte words, as specified by the previous field
+
+ This is the actual content of the compressed bitmap.
+
+ - 4-byte position of the current RLW for the compressed
+ bitmap
+
+All words are stored in network byte order for their corresponding
+sizes.
+
+The compressed bitmap is stored in a form of run-length encoding, as
+follows. It consists of a concatenation of an arbitrary number of
+chunks. Each chunk consists of one or more 64-bit words
+
+ H L_1 L_2 L_3 .... L_M
+
+H is called RLW (run length word). It consists of (from lower to higher
+order bits):
+
+ - 1 bit: the repeated bit B
+
+ - 32 bits: repetition count K (unsigned)
+
+ - 31 bits: literal word count M (unsigned)
+
+The bitstream represented by the above chunk is then:
+
+ - K repetitions of B
+
+ - The bits stored in `L_1` through `L_M`. Within a word, bits at
+ lower order come earlier in the stream than those at higher
+ order.
+
+The next word after `L_M` (if any) must again be a RLW, for the next
+chunk. For efficient appending to the bitstream, the EWAH stores a
+pointer to the last RLW in the stream.
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (8 preceding siblings ...)
2013-11-14 12:44 ` [PATCH v3 09/21] documentation: add documentation for the bitmap format Jeff King
@ 2013-11-14 12:44 ` Jeff King
2013-11-24 21:36 ` Thomas Rast
2013-11-29 21:21 ` Thomas Rast
2013-11-14 12:45 ` [PATCH v3 11/21] pack-objects: use bitmaps when packing objects Jeff King
` (11 subsequent siblings)
21 siblings, 2 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:44 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
A bitmap index is a `.bitmap` file that can be found inside
`$GIT_DIR/objects/pack/`, next to its corresponding packfile, and
contains precalculated reachability information for selected commits.
The full specification of the format for these bitmap indexes can be found
in `Documentation/technical/bitmap-format.txt`.
For a given commit SHA1, if it happens to be available in the bitmap
index, its bitmap will represent every single object that is reachable
from the commit itself. The nth bit in the bitmap is the nth object in
the packfile; if it's set to 1, the object is reachable.
By using the bitmaps available in the index, this commit implements
several new functions:
- `prepare_bitmap_git`
- `prepare_bitmap_walk`
- `traverse_bitmap_commit_list`
- `reuse_partial_packfile_from_bitmap`
The `prepare_bitmap_walk` function tries to build a bitmap of all the
objects that can be reached from the commit roots of a given `rev_info`
struct by using the following algorithm:
- If all the interesting commits for a revision walk are available in
the index, the resulting reachability bitmap is the bitwise OR of all
the individual bitmaps.
- When the full set of WANTs is not available in the index, we perform a
partial revision walk using the commits that don't have bitmaps as
roots, and limiting the revision walk as soon as we reach a commit that
has a corresponding bitmap. The earlier OR'ed bitmap with all the
indexed commits can now be completed as this walk progresses, so the end
result is the full reachability list.
- For revision walks with a HAVEs set (a set of commits that are deemed
uninteresting), first we perform the same method as for the WANTs, but
using our HAVEs as roots, in order to obtain a full reachability bitmap
of all the uninteresting commits. This bitmap then can be used to:
a) limit the subsequent walk when building the WANTs bitmap
b) finding the final set of interesting commits by performing an
AND-NOT of the WANTs and the HAVEs.
If `prepare_bitmap_walk` runs successfully, the resulting bitmap is
stored and the equivalent of a `traverse_commit_list` call can be
performed by using `traverse_bitmap_commit_list`; the bitmap version
of this call yields the objects straight from the packfile index
(without having to look them up or parse them) and hence is several
orders of magnitude faster.
As an extra optimization, when `prepare_bitmap_walk` succeeds, the
`reuse_partial_packfile_from_bitmap` call can be attempted: it will find
the amount of objects at the beginning of the on-disk packfile that can
be reused as-is, and return an offset into the packfile. The source
packfile can then be loaded and the bytes up to `offset` can be written
directly to the result without having to consider the entires inside the
packfile individually.
If the `prepare_bitmap_walk` call fails (e.g. because no bitmap files
are available), the `rev_info` struct is left untouched, and can be used
to perform a manual rev-walk using `traverse_commit_list`.
Hence, this new set of functions are a generic API that allows to
perform the equivalent of
git rev-list --objects [roots...] [^uninteresting...]
for any set of commits, even if they don't have specific bitmaps
generated for them.
In further patches, we'll use this bitmap traversal optimization to
speed up the `pack-objects` and `rev-list` commands.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Makefile | 2 +
khash.h | 338 ++++++++++++++++++++
pack-bitmap.c | 970 ++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
pack-bitmap.h | 43 +++
4 files changed, 1353 insertions(+)
create mode 100644 khash.h
create mode 100644 pack-bitmap.c
create mode 100644 pack-bitmap.h
diff --git a/Makefile b/Makefile
index 64a1ed7..b983d78 100644
--- a/Makefile
+++ b/Makefile
@@ -699,6 +699,7 @@ LIB_H += object.h
LIB_H += pack-objects.h
LIB_H += pack-revindex.h
LIB_H += pack.h
+LIB_H += pack-bitmap.h
LIB_H += parse-options.h
LIB_H += patch-ids.h
LIB_H += pathspec.h
@@ -837,6 +838,7 @@ LIB_OBJS += notes-cache.o
LIB_OBJS += notes-merge.o
LIB_OBJS += notes-utils.o
LIB_OBJS += object.o
+LIB_OBJS += pack-bitmap.o
LIB_OBJS += pack-check.o
LIB_OBJS += pack-objects.o
LIB_OBJS += pack-revindex.o
diff --git a/khash.h b/khash.h
new file mode 100644
index 0000000..57ff603
--- /dev/null
+++ b/khash.h
@@ -0,0 +1,338 @@
+/* The MIT License
+
+ Copyright (c) 2008, 2009, 2011 by Attractive Chaos <attractor@live.co.uk>
+
+ Permission is hereby granted, free of charge, to any person obtaining
+ a copy of this software and associated documentation files (the
+ "Software"), to deal in the Software without restriction, including
+ without limitation the rights to use, copy, modify, merge, publish,
+ distribute, sublicense, and/or sell copies of the Software, and to
+ permit persons to whom the Software is furnished to do so, subject to
+ the following conditions:
+
+ The above copyright notice and this permission notice shall be
+ included in all copies or substantial portions of the Software.
+
+ THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
+ EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
+ NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
+ BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
+ ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
+ CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+ SOFTWARE.
+*/
+
+#ifndef __AC_KHASH_H
+#define __AC_KHASH_H
+
+#define AC_VERSION_KHASH_H "0.2.8"
+
+typedef uint32_t khint32_t;
+typedef uint64_t khint64_t;
+
+typedef khint32_t khint_t;
+typedef khint_t khiter_t;
+
+#define __ac_isempty(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&2)
+#define __ac_isdel(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&1)
+#define __ac_iseither(flag, i) ((flag[i>>4]>>((i&0xfU)<<1))&3)
+#define __ac_set_isdel_false(flag, i) (flag[i>>4]&=~(1ul<<((i&0xfU)<<1)))
+#define __ac_set_isempty_false(flag, i) (flag[i>>4]&=~(2ul<<((i&0xfU)<<1)))
+#define __ac_set_isboth_false(flag, i) (flag[i>>4]&=~(3ul<<((i&0xfU)<<1)))
+#define __ac_set_isdel_true(flag, i) (flag[i>>4]|=1ul<<((i&0xfU)<<1))
+
+#define __ac_fsize(m) ((m) < 16? 1 : (m)>>4)
+
+#define kroundup32(x) (--(x), (x)|=(x)>>1, (x)|=(x)>>2, (x)|=(x)>>4, (x)|=(x)>>8, (x)|=(x)>>16, ++(x))
+
+static inline khint_t __ac_X31_hash_string(const char *s)
+{
+ khint_t h = (khint_t)*s;
+ if (h) for (++s ; *s; ++s) h = (h << 5) - h + (khint_t)*s;
+ return h;
+}
+
+#define kh_str_hash_func(key) __ac_X31_hash_string(key)
+#define kh_str_hash_equal(a, b) (strcmp(a, b) == 0)
+
+static const double __ac_HASH_UPPER = 0.77;
+
+#define __KHASH_TYPE(name, khkey_t, khval_t) \
+ typedef struct { \
+ khint_t n_buckets, size, n_occupied, upper_bound; \
+ khint32_t *flags; \
+ khkey_t *keys; \
+ khval_t *vals; \
+ } kh_##name##_t;
+
+#define __KHASH_PROTOTYPES(name, khkey_t, khval_t) \
+ extern kh_##name##_t *kh_init_##name(void); \
+ extern void kh_destroy_##name(kh_##name##_t *h); \
+ extern void kh_clear_##name(kh_##name##_t *h); \
+ extern khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key); \
+ extern int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets); \
+ extern khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret); \
+ extern void kh_del_##name(kh_##name##_t *h, khint_t x);
+
+#define __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
+ SCOPE kh_##name##_t *kh_init_##name(void) { \
+ return (kh_##name##_t*)xcalloc(1, sizeof(kh_##name##_t)); \
+ } \
+ SCOPE void kh_destroy_##name(kh_##name##_t *h) \
+ { \
+ if (h) { \
+ free((void *)h->keys); free(h->flags); \
+ free((void *)h->vals); \
+ free(h); \
+ } \
+ } \
+ SCOPE void kh_clear_##name(kh_##name##_t *h) \
+ { \
+ if (h && h->flags) { \
+ memset(h->flags, 0xaa, __ac_fsize(h->n_buckets) * sizeof(khint32_t)); \
+ h->size = h->n_occupied = 0; \
+ } \
+ } \
+ SCOPE khint_t kh_get_##name(const kh_##name##_t *h, khkey_t key) \
+ { \
+ if (h->n_buckets) { \
+ khint_t k, i, last, mask, step = 0; \
+ mask = h->n_buckets - 1; \
+ k = __hash_func(key); i = k & mask; \
+ last = i; \
+ while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \
+ i = (i + (++step)) & mask; \
+ if (i == last) return h->n_buckets; \
+ } \
+ return __ac_iseither(h->flags, i)? h->n_buckets : i; \
+ } else return 0; \
+ } \
+ SCOPE int kh_resize_##name(kh_##name##_t *h, khint_t new_n_buckets) \
+ { /* This function uses 0.25*n_buckets bytes of working space instead of [sizeof(key_t+val_t)+.25]*n_buckets. */ \
+ khint32_t *new_flags = NULL; \
+ khint_t j = 1; \
+ { \
+ kroundup32(new_n_buckets); \
+ if (new_n_buckets < 4) new_n_buckets = 4; \
+ if (h->size >= (khint_t)(new_n_buckets * __ac_HASH_UPPER + 0.5)) j = 0; /* requested size is too small */ \
+ else { /* hash table size to be changed (shrink or expand); rehash */ \
+ new_flags = (khint32_t*)xmalloc(__ac_fsize(new_n_buckets) * sizeof(khint32_t)); \
+ if (!new_flags) return -1; \
+ memset(new_flags, 0xaa, __ac_fsize(new_n_buckets) * sizeof(khint32_t)); \
+ if (h->n_buckets < new_n_buckets) { /* expand */ \
+ khkey_t *new_keys = (khkey_t*)xrealloc((void *)h->keys, new_n_buckets * sizeof(khkey_t)); \
+ if (!new_keys) return -1; \
+ h->keys = new_keys; \
+ if (kh_is_map) { \
+ khval_t *new_vals = (khval_t*)xrealloc((void *)h->vals, new_n_buckets * sizeof(khval_t)); \
+ if (!new_vals) return -1; \
+ h->vals = new_vals; \
+ } \
+ } /* otherwise shrink */ \
+ } \
+ } \
+ if (j) { /* rehashing is needed */ \
+ for (j = 0; j != h->n_buckets; ++j) { \
+ if (__ac_iseither(h->flags, j) == 0) { \
+ khkey_t key = h->keys[j]; \
+ khval_t val; \
+ khint_t new_mask; \
+ new_mask = new_n_buckets - 1; \
+ if (kh_is_map) val = h->vals[j]; \
+ __ac_set_isdel_true(h->flags, j); \
+ while (1) { /* kick-out process; sort of like in Cuckoo hashing */ \
+ khint_t k, i, step = 0; \
+ k = __hash_func(key); \
+ i = k & new_mask; \
+ while (!__ac_isempty(new_flags, i)) i = (i + (++step)) & new_mask; \
+ __ac_set_isempty_false(new_flags, i); \
+ if (i < h->n_buckets && __ac_iseither(h->flags, i) == 0) { /* kick out the existing element */ \
+ { khkey_t tmp = h->keys[i]; h->keys[i] = key; key = tmp; } \
+ if (kh_is_map) { khval_t tmp = h->vals[i]; h->vals[i] = val; val = tmp; } \
+ __ac_set_isdel_true(h->flags, i); /* mark it as deleted in the old hash table */ \
+ } else { /* write the element and jump out of the loop */ \
+ h->keys[i] = key; \
+ if (kh_is_map) h->vals[i] = val; \
+ break; \
+ } \
+ } \
+ } \
+ } \
+ if (h->n_buckets > new_n_buckets) { /* shrink the hash table */ \
+ h->keys = (khkey_t*)xrealloc((void *)h->keys, new_n_buckets * sizeof(khkey_t)); \
+ if (kh_is_map) h->vals = (khval_t*)xrealloc((void *)h->vals, new_n_buckets * sizeof(khval_t)); \
+ } \
+ free(h->flags); /* free the working space */ \
+ h->flags = new_flags; \
+ h->n_buckets = new_n_buckets; \
+ h->n_occupied = h->size; \
+ h->upper_bound = (khint_t)(h->n_buckets * __ac_HASH_UPPER + 0.5); \
+ } \
+ return 0; \
+ } \
+ SCOPE khint_t kh_put_##name(kh_##name##_t *h, khkey_t key, int *ret) \
+ { \
+ khint_t x; \
+ if (h->n_occupied >= h->upper_bound) { /* update the hash table */ \
+ if (h->n_buckets > (h->size<<1)) { \
+ if (kh_resize_##name(h, h->n_buckets - 1) < 0) { /* clear "deleted" elements */ \
+ *ret = -1; return h->n_buckets; \
+ } \
+ } else if (kh_resize_##name(h, h->n_buckets + 1) < 0) { /* expand the hash table */ \
+ *ret = -1; return h->n_buckets; \
+ } \
+ } /* TODO: to implement automatically shrinking; resize() already support shrinking */ \
+ { \
+ khint_t k, i, site, last, mask = h->n_buckets - 1, step = 0; \
+ x = site = h->n_buckets; k = __hash_func(key); i = k & mask; \
+ if (__ac_isempty(h->flags, i)) x = i; /* for speed up */ \
+ else { \
+ last = i; \
+ while (!__ac_isempty(h->flags, i) && (__ac_isdel(h->flags, i) || !__hash_equal(h->keys[i], key))) { \
+ if (__ac_isdel(h->flags, i)) site = i; \
+ i = (i + (++step)) & mask; \
+ if (i == last) { x = site; break; } \
+ } \
+ if (x == h->n_buckets) { \
+ if (__ac_isempty(h->flags, i) && site != h->n_buckets) x = site; \
+ else x = i; \
+ } \
+ } \
+ } \
+ if (__ac_isempty(h->flags, x)) { /* not present at all */ \
+ h->keys[x] = key; \
+ __ac_set_isboth_false(h->flags, x); \
+ ++h->size; ++h->n_occupied; \
+ *ret = 1; \
+ } else if (__ac_isdel(h->flags, x)) { /* deleted */ \
+ h->keys[x] = key; \
+ __ac_set_isboth_false(h->flags, x); \
+ ++h->size; \
+ *ret = 2; \
+ } else *ret = 0; /* Don't touch h->keys[x] if present and not deleted */ \
+ return x; \
+ } \
+ SCOPE void kh_del_##name(kh_##name##_t *h, khint_t x) \
+ { \
+ if (x != h->n_buckets && !__ac_iseither(h->flags, x)) { \
+ __ac_set_isdel_true(h->flags, x); \
+ --h->size; \
+ } \
+ }
+
+#define KHASH_DECLARE(name, khkey_t, khval_t) \
+ __KHASH_TYPE(name, khkey_t, khval_t) \
+ __KHASH_PROTOTYPES(name, khkey_t, khval_t)
+
+#define KHASH_INIT2(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
+ __KHASH_TYPE(name, khkey_t, khval_t) \
+ __KHASH_IMPL(name, SCOPE, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal)
+
+#define KHASH_INIT(name, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal) \
+ KHASH_INIT2(name, static inline, khkey_t, khval_t, kh_is_map, __hash_func, __hash_equal)
+
+/* Other convenient macros... */
+
+/*! @function
+ @abstract Test whether a bucket contains data.
+ @param h Pointer to the hash table [khash_t(name)*]
+ @param x Iterator to the bucket [khint_t]
+ @return 1 if containing data; 0 otherwise [int]
+ */
+#define kh_exist(h, x) (!__ac_iseither((h)->flags, (x)))
+
+/*! @function
+ @abstract Get key given an iterator
+ @param h Pointer to the hash table [khash_t(name)*]
+ @param x Iterator to the bucket [khint_t]
+ @return Key [type of keys]
+ */
+#define kh_key(h, x) ((h)->keys[x])
+
+/*! @function
+ @abstract Get value given an iterator
+ @param h Pointer to the hash table [khash_t(name)*]
+ @param x Iterator to the bucket [khint_t]
+ @return Value [type of values]
+ @discussion For hash sets, calling this results in segfault.
+ */
+#define kh_val(h, x) ((h)->vals[x])
+
+/*! @function
+ @abstract Alias of kh_val()
+ */
+#define kh_value(h, x) ((h)->vals[x])
+
+/*! @function
+ @abstract Get the start iterator
+ @param h Pointer to the hash table [khash_t(name)*]
+ @return The start iterator [khint_t]
+ */
+#define kh_begin(h) (khint_t)(0)
+
+/*! @function
+ @abstract Get the end iterator
+ @param h Pointer to the hash table [khash_t(name)*]
+ @return The end iterator [khint_t]
+ */
+#define kh_end(h) ((h)->n_buckets)
+
+/*! @function
+ @abstract Get the number of elements in the hash table
+ @param h Pointer to the hash table [khash_t(name)*]
+ @return Number of elements in the hash table [khint_t]
+ */
+#define kh_size(h) ((h)->size)
+
+/*! @function
+ @abstract Get the number of buckets in the hash table
+ @param h Pointer to the hash table [khash_t(name)*]
+ @return Number of buckets in the hash table [khint_t]
+ */
+#define kh_n_buckets(h) ((h)->n_buckets)
+
+/*! @function
+ @abstract Iterate over the entries in the hash table
+ @param h Pointer to the hash table [khash_t(name)*]
+ @param kvar Variable to which key will be assigned
+ @param vvar Variable to which value will be assigned
+ @param code Block of code to execute
+ */
+#define kh_foreach(h, kvar, vvar, code) { khint_t __i; \
+ for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \
+ if (!kh_exist(h,__i)) continue; \
+ (kvar) = kh_key(h,__i); \
+ (vvar) = kh_val(h,__i); \
+ code; \
+ } }
+
+/*! @function
+ @abstract Iterate over the values in the hash table
+ @param h Pointer to the hash table [khash_t(name)*]
+ @param vvar Variable to which value will be assigned
+ @param code Block of code to execute
+ */
+#define kh_foreach_value(h, vvar, code) { khint_t __i; \
+ for (__i = kh_begin(h); __i != kh_end(h); ++__i) { \
+ if (!kh_exist(h,__i)) continue; \
+ (vvar) = kh_val(h,__i); \
+ code; \
+ } }
+
+static inline khint_t __kh_oid_hash(const unsigned char *oid)
+{
+ khint_t hash;
+ memcpy(&hash, oid, sizeof(hash));
+ return hash;
+}
+
+#define __kh_oid_cmp(a, b) (hashcmp(a, b) == 0)
+
+KHASH_INIT(sha1, const unsigned char *, void *, 1, __kh_oid_hash, __kh_oid_cmp)
+typedef kh_sha1_t khash_sha1;
+
+KHASH_INIT(sha1_pos, const unsigned char *, int, 1, __kh_oid_hash, __kh_oid_cmp)
+typedef kh_sha1_pos_t khash_sha1_pos;
+
+#endif /* __AC_KHASH_H */
diff --git a/pack-bitmap.c b/pack-bitmap.c
new file mode 100644
index 0000000..340f02b
--- /dev/null
+++ b/pack-bitmap.c
@@ -0,0 +1,970 @@
+#include "cache.h"
+#include "commit.h"
+#include "tag.h"
+#include "diff.h"
+#include "revision.h"
+#include "progress.h"
+#include "list-objects.h"
+#include "pack.h"
+#include "pack-bitmap.h"
+#include "pack-revindex.h"
+#include "pack-objects.h"
+
+/*
+ * An entry on the bitmap index, representing the bitmap for a given
+ * commit.
+ */
+struct stored_bitmap {
+ unsigned char sha1[20];
+ struct ewah_bitmap *root;
+ struct stored_bitmap *xor;
+ int flags;
+};
+
+/*
+ * The currently active bitmap index. By design, repositories only have
+ * a single bitmap index available (the index for the biggest packfile in
+ * the repository), since bitmap indexes need full closure.
+ *
+ * If there is more than one bitmap index available (e.g. because of alternates),
+ * the active bitmap index is the largest one.
+ */
+static struct bitmap_index {
+ /* Packfile to which this bitmap index belongs to */
+ struct packed_git *pack;
+
+ /* reverse index for the packfile */
+ struct pack_revindex *reverse_index;
+
+ /*
+ * Mark the first `reuse_objects` in the packfile as reused:
+ * they will be sent as-is without using them for repacking
+ * calculations
+ */
+ uint32_t reuse_objects;
+
+ /* mmapped buffer of the whole bitmap index */
+ unsigned char *map;
+ size_t map_size; /* size of the mmaped buffer */
+ size_t map_pos; /* current position when loading the index */
+
+ /*
+ * Type indexes.
+ *
+ * Each bitmap marks which objects in the packfile are of the given
+ * type. This provides type information when yielding the objects from
+ * the packfile during a walk, which allows for better delta bases.
+ */
+ struct ewah_bitmap *commits;
+ struct ewah_bitmap *trees;
+ struct ewah_bitmap *blobs;
+ struct ewah_bitmap *tags;
+
+ /* Map from SHA1 -> `stored_bitmap` for all the bitmapped comits */
+ khash_sha1 *bitmaps;
+
+ /* Number of bitmapped commits */
+ uint32_t entry_count;
+
+ /*
+ * Extended index.
+ *
+ * When trying to perform bitmap operations with objects that are not
+ * packed in `pack`, these objects are added to this "fake index" and
+ * are assumed to appear at the end of the packfile for all operations
+ */
+ struct eindex {
+ struct object **objects;
+ uint32_t *hashes;
+ uint32_t count, alloc;
+ khash_sha1_pos *positions;
+ } ext_index;
+
+ /* Bitmap result of the last performed walk */
+ struct bitmap *result;
+
+ /* Version of the bitmap index */
+ unsigned int version;
+
+ unsigned loaded : 1;
+
+} bitmap_git;
+
+static struct ewah_bitmap *lookup_stored_bitmap(struct stored_bitmap *st)
+{
+ struct ewah_bitmap *parent;
+ struct ewah_bitmap *composed;
+
+ if (st->xor == NULL)
+ return st->root;
+
+ composed = ewah_pool_new();
+ parent = lookup_stored_bitmap(st->xor);
+ ewah_xor(st->root, parent, composed);
+
+ ewah_pool_free(st->root);
+ st->root = composed;
+ st->xor = NULL;
+
+ return composed;
+}
+
+/*
+ * Read a bitmap from the current read position on the mmaped
+ * index, and increase the read position accordingly
+ */
+static struct ewah_bitmap *read_bitmap_1(struct bitmap_index *index)
+{
+ struct ewah_bitmap *b = ewah_pool_new();
+
+ int bitmap_size = ewah_read_mmap(b,
+ index->map + index->map_pos,
+ index->map_size - index->map_pos);
+
+ if (bitmap_size < 0) {
+ error("Failed to load bitmap index (corrupted?)");
+ ewah_pool_free(b);
+ return NULL;
+ }
+
+ index->map_pos += bitmap_size;
+ return b;
+}
+
+static int load_bitmap_header(struct bitmap_index *index)
+{
+ struct bitmap_disk_header *header = (void *)index->map;
+
+ if (index->map_size < sizeof(*header) + 20)
+ return error("Corrupted bitmap index (missing header data)");
+
+ if (memcmp(header->magic, BITMAP_IDX_SIGNATURE, sizeof(BITMAP_IDX_SIGNATURE)) != 0)
+ return error("Corrupted bitmap index file (wrong header)");
+
+ index->version = ntohs(header->version);
+ if (index->version != 1)
+ return error("Unsupported version for bitmap index file (%d)", index->version);
+
+ /* Parse known bitmap format options */
+ {
+ uint32_t flags = ntohs(header->options);
+
+ if ((flags & BITMAP_OPT_FULL_DAG) == 0)
+ return error("Unsupported options for bitmap index file "
+ "(Git requires BITMAP_OPT_FULL_DAG)");
+ }
+
+ index->entry_count = ntohl(header->entry_count);
+ index->map_pos += sizeof(*header);
+ return 0;
+}
+
+static struct stored_bitmap *store_bitmap(struct bitmap_index *index,
+ struct ewah_bitmap *root,
+ const unsigned char *sha1,
+ struct stored_bitmap *xor_with,
+ int flags)
+{
+ struct stored_bitmap *stored;
+ khiter_t hash_pos;
+ int ret;
+
+ stored = xmalloc(sizeof(struct stored_bitmap));
+ stored->root = root;
+ stored->xor = xor_with;
+ stored->flags = flags;
+ hashcpy(stored->sha1, sha1);
+
+ hash_pos = kh_put_sha1(index->bitmaps, stored->sha1, &ret);
+
+ /* a 0 return code means the insertion succeeded with no changes,
+ * because the SHA1 already existed on the map. this is bad, there
+ * shouldn't be duplicated commits in the index */
+ if (ret == 0) {
+ error("Duplicate entry in bitmap index: %s", sha1_to_hex(sha1));
+ return NULL;
+ }
+
+ kh_value(index->bitmaps, hash_pos) = stored;
+ return stored;
+}
+
+static int load_bitmap_entries_v1(struct bitmap_index *index)
+{
+ static const size_t MAX_XOR_OFFSET = 160;
+
+ uint32_t i;
+ struct stored_bitmap **recent_bitmaps;
+ struct bitmap_disk_entry *entry;
+
+ recent_bitmaps = xcalloc(MAX_XOR_OFFSET, sizeof(struct stored_bitmap));
+
+ for (i = 0; i < index->entry_count; ++i) {
+ int xor_offset, flags;
+ struct ewah_bitmap *bitmap = NULL;
+ struct stored_bitmap *xor_bitmap = NULL;
+ uint32_t commit_idx_pos;
+ const unsigned char *sha1;
+
+ entry = (struct bitmap_disk_entry *)(index->map + index->map_pos);
+ index->map_pos += sizeof(struct bitmap_disk_entry);
+
+ commit_idx_pos = ntohl(entry->object_pos);
+ sha1 = nth_packed_object_sha1(index->pack, commit_idx_pos);
+
+ xor_offset = (int)entry->xor_offset;
+ flags = (int)entry->flags;
+
+ bitmap = read_bitmap_1(index);
+ if (!bitmap)
+ return -1;
+
+ if (xor_offset > MAX_XOR_OFFSET || xor_offset > i)
+ return error("Corrupted bitmap pack index");
+
+ if (xor_offset > 0) {
+ xor_bitmap = recent_bitmaps[(i - xor_offset) % MAX_XOR_OFFSET];
+
+ if (xor_bitmap == NULL)
+ return error("Invalid XOR offset in bitmap pack index");
+ }
+
+ recent_bitmaps[i % MAX_XOR_OFFSET] = store_bitmap(
+ index, bitmap, sha1, xor_bitmap, flags);
+ }
+
+ return 0;
+}
+
+static int open_pack_bitmap_1(struct packed_git *packfile)
+{
+ int fd;
+ struct stat st;
+ char *idx_name;
+
+ if (open_pack_index(packfile))
+ return -1;
+
+ idx_name = pack_bitmap_filename(packfile);
+ fd = git_open_noatime(idx_name);
+ free(idx_name);
+
+ if (fd < 0)
+ return -1;
+
+ if (fstat(fd, &st)) {
+ close(fd);
+ return -1;
+ }
+
+ if (bitmap_git.pack) {
+ warning("ignoring extra bitmap file: %s", idx_name);
+ close(fd);
+ return -1;
+ }
+
+ bitmap_git.pack = packfile;
+ bitmap_git.map_size = xsize_t(st.st_size);
+ bitmap_git.map = xmmap(NULL, bitmap_git.map_size, PROT_READ, MAP_PRIVATE, fd, 0);
+ bitmap_git.map_pos = 0;
+ close(fd);
+
+ if (load_bitmap_header(&bitmap_git) < 0) {
+ munmap(bitmap_git.map, bitmap_git.map_size);
+ bitmap_git.map = NULL;
+ bitmap_git.map_size = 0;
+ return -1;
+ }
+
+ return 0;
+}
+
+static int load_pack_bitmap(void)
+{
+ assert(bitmap_git.map && !bitmap_git.loaded);
+
+ bitmap_git.bitmaps = kh_init_sha1();
+ bitmap_git.ext_index.positions = kh_init_sha1_pos();
+ bitmap_git.reverse_index = revindex_for_pack(bitmap_git.pack);
+
+ if (!(bitmap_git.commits = read_bitmap_1(&bitmap_git)) ||
+ !(bitmap_git.trees = read_bitmap_1(&bitmap_git)) ||
+ !(bitmap_git.blobs = read_bitmap_1(&bitmap_git)) ||
+ !(bitmap_git.tags = read_bitmap_1(&bitmap_git)))
+ goto failed;
+
+ if (load_bitmap_entries_v1(&bitmap_git) < 0)
+ goto failed;
+
+ bitmap_git.loaded = 1;
+ return 0;
+
+failed:
+ munmap(bitmap_git.map, bitmap_git.map_size);
+ bitmap_git.map = NULL;
+ bitmap_git.map_size = 0;
+ return -1;
+}
+
+char *pack_bitmap_filename(struct packed_git *p)
+{
+ char *idx_name;
+ int len;
+
+ len = strlen(p->pack_name) - strlen(".pack");
+ idx_name = xmalloc(len + strlen(".bitmap") + 1);
+
+ memcpy(idx_name, p->pack_name, len);
+ memcpy(idx_name + len, ".bitmap", strlen(".bitmap") + 1);
+
+ return idx_name;
+}
+
+static int open_pack_bitmap(void)
+{
+ struct packed_git *p;
+ int ret = -1;
+
+ assert(!bitmap_git.map && !bitmap_git.loaded);
+
+ prepare_packed_git();
+ for (p = packed_git; p; p = p->next) {
+ if (open_pack_bitmap_1(p) == 0)
+ ret = 0;
+ }
+
+ return ret;
+}
+
+int prepare_bitmap_git(void)
+{
+ if (bitmap_git.loaded)
+ return 0;
+
+ if (!open_pack_bitmap())
+ return load_pack_bitmap();
+
+ return -1;
+}
+
+struct include_data {
+ struct bitmap *base;
+ struct bitmap *seen;
+};
+
+static inline int bitmap_position_extended(const unsigned char *sha1)
+{
+ khash_sha1_pos *positions = bitmap_git.ext_index.positions;
+ khiter_t pos = kh_get_sha1_pos(positions, sha1);
+
+ if (pos < kh_end(positions)) {
+ int bitmap_pos = kh_value(positions, pos);
+ return bitmap_pos + bitmap_git.pack->num_objects;
+ }
+
+ return -1;
+}
+
+static inline int bitmap_position_packfile(const unsigned char *sha1)
+{
+ off_t offset = find_pack_entry_one(sha1, bitmap_git.pack);
+ if (!offset)
+ return -1;
+
+ return find_revindex_position(bitmap_git.reverse_index, offset);
+}
+
+static int bitmap_position(const unsigned char *sha1)
+{
+ int pos = bitmap_position_packfile(sha1);
+ return (pos >= 0) ? pos : bitmap_position_extended(sha1);
+}
+
+static int ext_index_add_object(struct object *object, const char *name)
+{
+ struct eindex *eindex = &bitmap_git.ext_index;
+
+ khiter_t hash_pos;
+ int hash_ret;
+ int bitmap_pos;
+
+ hash_pos = kh_put_sha1_pos(eindex->positions, object->sha1, &hash_ret);
+ if (hash_ret > 0) {
+ if (eindex->count >= eindex->alloc) {
+ eindex->alloc = (eindex->alloc + 16) * 3 / 2;
+ eindex->objects = xrealloc(eindex->objects,
+ eindex->alloc * sizeof(struct object *));
+ eindex->hashes = xrealloc(eindex->hashes,
+ eindex->alloc * sizeof(uint32_t));
+ }
+
+ bitmap_pos = eindex->count;
+ eindex->objects[eindex->count] = object;
+ eindex->hashes[eindex->count] = pack_name_hash(name);
+ kh_value(eindex->positions, hash_pos) = bitmap_pos;
+ eindex->count++;
+ } else {
+ bitmap_pos = kh_value(eindex->positions, hash_pos);
+ }
+
+ return bitmap_pos + bitmap_git.pack->num_objects;
+}
+
+static void show_object(struct object *object, const struct name_path *path,
+ const char *last, void *data)
+{
+ struct bitmap *base = data;
+ int bitmap_pos;
+
+ bitmap_pos = bitmap_position(object->sha1);
+
+ if (bitmap_pos < 0) {
+ char *name = path_name(path, last);
+ bitmap_pos = ext_index_add_object(object, name);
+ free(name);
+ }
+
+ bitmap_set(base, bitmap_pos);
+}
+
+static void show_commit(struct commit *commit, void *data)
+{
+}
+
+static int add_to_include_set(struct include_data *data,
+ const unsigned char *sha1,
+ int bitmap_pos)
+{
+ khiter_t hash_pos;
+
+ if (data->seen && bitmap_get(data->seen, bitmap_pos))
+ return 0;
+
+ if (bitmap_get(data->base, bitmap_pos))
+ return 0;
+
+ hash_pos = kh_get_sha1(bitmap_git.bitmaps, sha1);
+ if (hash_pos < kh_end(bitmap_git.bitmaps)) {
+ struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, hash_pos);
+ bitmap_or_ewah(data->base, lookup_stored_bitmap(st));
+ return 0;
+ }
+
+ bitmap_set(data->base, bitmap_pos);
+ return 1;
+}
+
+static int should_include(struct commit *commit, void *_data)
+{
+ struct include_data *data = _data;
+ int bitmap_pos;
+
+ bitmap_pos = bitmap_position(commit->object.sha1);
+ if (bitmap_pos < 0)
+ bitmap_pos = ext_index_add_object((struct object *)commit, NULL);
+
+ if (!add_to_include_set(data, commit->object.sha1, bitmap_pos)) {
+ struct commit_list *parent = commit->parents;
+
+ while (parent) {
+ parent->item->object.flags |= SEEN;
+ parent = parent->next;
+ }
+
+ return 0;
+ }
+
+ return 1;
+}
+
+static struct bitmap *find_objects(struct rev_info *revs,
+ struct object_list *roots,
+ struct bitmap *seen)
+{
+ struct bitmap *base = NULL;
+ int needs_walk = 0;
+
+ struct object_list *not_mapped = NULL;
+
+ /*
+ * Go through all the roots for the walk. The ones that have bitmaps
+ * on the bitmap index will be `or`ed together to form an initial
+ * global reachability analysis.
+ *
+ * The ones without bitmaps in the index will be stored in the
+ * `not_mapped_list` for further processing.
+ */
+ while (roots) {
+ struct object *object = roots->item;
+ roots = roots->next;
+
+ if (object->type == OBJ_COMMIT) {
+ khiter_t pos = kh_get_sha1(bitmap_git.bitmaps, object->sha1);
+
+ if (pos < kh_end(bitmap_git.bitmaps)) {
+ struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, pos);
+ struct ewah_bitmap *or_with = lookup_stored_bitmap(st);
+
+ if (base == NULL)
+ base = ewah_to_bitmap(or_with);
+ else
+ bitmap_or_ewah(base, or_with);
+
+ object->flags |= SEEN;
+ continue;
+ }
+ }
+
+ object_list_insert(object, ¬_mapped);
+ }
+
+ /*
+ * Best case scenario: We found bitmaps for all the roots,
+ * so the resulting `or` bitmap has the full reachability analysis
+ */
+ if (not_mapped == NULL)
+ return base;
+
+ roots = not_mapped;
+
+ /*
+ * Let's iterate through all the roots that don't have bitmaps to
+ * check if we can determine them to be reachable from the existing
+ * global bitmap.
+ *
+ * If we cannot find them in the existing global bitmap, we'll need
+ * to push them to an actual walk and run it until we can confirm
+ * they are reachable
+ */
+ while (roots) {
+ struct object *object = roots->item;
+ int pos;
+
+ roots = roots->next;
+ pos = bitmap_position(object->sha1);
+
+ if (pos < 0 || base == NULL || !bitmap_get(base, pos)) {
+ object->flags &= ~UNINTERESTING;
+ add_pending_object(revs, object, "");
+ needs_walk = 1;
+ } else {
+ object->flags |= SEEN;
+ }
+ }
+
+ if (needs_walk) {
+ struct include_data incdata;
+
+ if (base == NULL)
+ base = bitmap_new();
+
+ incdata.base = base;
+ incdata.seen = seen;
+
+ revs->include_check = should_include;
+ revs->include_check_data = &incdata;
+
+ if (prepare_revision_walk(revs))
+ die("revision walk setup failed");
+
+ traverse_commit_list(revs, show_commit, show_object, base);
+ }
+
+ return base;
+}
+
+static void show_extended_objects(struct bitmap *objects,
+ show_reachable_fn show_reach)
+{
+ struct eindex *eindex = &bitmap_git.ext_index;
+ uint32_t i;
+
+ for (i = 0; i < eindex->count; ++i) {
+ struct object *obj;
+
+ if (!bitmap_get(objects, bitmap_git.pack->num_objects + i))
+ continue;
+
+ obj = eindex->objects[i];
+ show_reach(obj->sha1, obj->type, 0, eindex->hashes[i], NULL, 0);
+ }
+}
+
+static void show_objects_for_type(
+ struct bitmap *objects,
+ struct ewah_bitmap *type_filter,
+ enum object_type object_type,
+ show_reachable_fn show_reach)
+{
+ size_t pos = 0, i = 0;
+ uint32_t offset;
+
+ struct ewah_iterator it;
+ eword_t filter;
+
+ if (bitmap_git.reuse_objects == bitmap_git.pack->num_objects)
+ return;
+
+ ewah_iterator_init(&it, type_filter);
+
+ while (i < objects->word_alloc && ewah_iterator_next(&filter, &it)) {
+ eword_t word = objects->words[i] & filter;
+
+ for (offset = 0; offset < BITS_IN_WORD; ++offset) {
+ const unsigned char *sha1;
+ struct revindex_entry *entry;
+ uint32_t hash = 0;
+
+ if ((word >> offset) == 0)
+ break;
+
+ offset += ewah_bit_ctz64(word >> offset);
+
+ if (pos + offset < bitmap_git.reuse_objects)
+ continue;
+
+ entry = &bitmap_git.reverse_index->revindex[pos + offset];
+ sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
+
+ show_reach(sha1, object_type, 0, hash, bitmap_git.pack, entry->offset);
+ }
+
+ pos += BITS_IN_WORD;
+ i++;
+ }
+}
+
+static int in_bitmapped_pack(struct object_list *roots)
+{
+ while (roots) {
+ struct object *object = roots->item;
+ roots = roots->next;
+
+ if (find_pack_entry_one(object->sha1, bitmap_git.pack) > 0)
+ return 1;
+ }
+
+ return 0;
+}
+
+int prepare_bitmap_walk(struct rev_info *revs)
+{
+ unsigned int i;
+ unsigned int pending_nr = revs->pending.nr;
+ struct object_array_entry *pending_e = revs->pending.objects;
+
+ struct object_list *wants = NULL;
+ struct object_list *haves = NULL;
+
+ struct bitmap *wants_bitmap = NULL;
+ struct bitmap *haves_bitmap = NULL;
+
+ if (!bitmap_git.loaded) {
+ /* try to open a bitmapped pack, but don't parse it yet
+ * because we may not need to use it */
+ if (open_pack_bitmap() < 0)
+ return -1;
+ }
+
+ for (i = 0; i < pending_nr; ++i) {
+ struct object *object = pending_e[i].item;
+
+ if (object->type == OBJ_NONE)
+ parse_object_or_die(object->sha1, NULL);
+
+ while (object->type == OBJ_TAG) {
+ struct tag *tag = (struct tag *) object;
+
+ if (object->flags & UNINTERESTING)
+ object_list_insert(object, &haves);
+ else
+ object_list_insert(object, &wants);
+
+ if (!tag->tagged)
+ die("bad tag");
+ object = parse_object_or_die(tag->tagged->sha1, NULL);
+ }
+
+ if (object->flags & UNINTERESTING)
+ object_list_insert(object, &haves);
+ else
+ object_list_insert(object, &wants);
+ }
+
+ /*
+ * if we have a HAVES list, but none of those haves is contained
+ * in the packfile that has a bitmap, we don't have anything to
+ * optimize here
+ */
+ if (haves && !in_bitmapped_pack(haves))
+ return -1;
+
+ /* if we don't want anything, we're done here */
+ if (!wants)
+ return -1;
+
+ /*
+ * now we're going to use bitmaps, so load the actual bitmap entries
+ * from disk. this is the point of no return; after this the rev_list
+ * becomes invalidated and we must perform the revwalk through bitmaps
+ */
+ if (!bitmap_git.loaded && load_pack_bitmap() < 0)
+ return -1;
+
+ revs->pending.nr = 0;
+ revs->pending.alloc = 0;
+ revs->pending.objects = NULL;
+
+ if (haves) {
+ haves_bitmap = find_objects(revs, haves, NULL);
+ reset_revision_walk();
+
+ if (haves_bitmap == NULL)
+ die("BUG: failed to perform bitmap walk");
+ }
+
+ wants_bitmap = find_objects(revs, wants, haves_bitmap);
+
+ if (!wants_bitmap)
+ die("BUG: failed to perform bitmap walk");
+
+ if (haves_bitmap)
+ bitmap_and_not(wants_bitmap, haves_bitmap);
+
+ bitmap_git.result = wants_bitmap;
+
+ bitmap_free(haves_bitmap);
+ return 0;
+}
+
+int reuse_partial_packfile_from_bitmap(struct packed_git **packfile,
+ uint32_t *entries,
+ off_t *up_to)
+{
+ /*
+ * Reuse the packfile content if we need more than
+ * 90% of its objects
+ */
+ static const double REUSE_PERCENT = 0.9;
+
+ struct bitmap *result = bitmap_git.result;
+ uint32_t reuse_threshold;
+ uint32_t i, reuse_objects = 0;
+
+ assert(result);
+
+ for (i = 0; i < result->word_alloc; ++i) {
+ if (result->words[i] != (eword_t)~0) {
+ reuse_objects += ewah_bit_ctz64(~result->words[i]);
+ break;
+ }
+
+ reuse_objects += BITS_IN_WORD;
+ }
+
+#ifdef GIT_BITMAP_DEBUG
+ {
+ const unsigned char *sha1;
+ struct revindex_entry *entry;
+
+ entry = &bitmap_git.reverse_index->revindex[reuse_objects];
+ sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
+
+ fprintf(stderr, "Failed to reuse at %d (%016llx)\n",
+ reuse_objects, result->words[i]);
+ fprintf(stderr, " %s\n", sha1_to_hex(sha1));
+ }
+#endif
+
+ if (!reuse_objects)
+ return -1;
+
+ if (reuse_objects >= bitmap_git.pack->num_objects) {
+ bitmap_git.reuse_objects = *entries = bitmap_git.pack->num_objects;
+ *up_to = -1; /* reuse the full pack */
+ *packfile = bitmap_git.pack;
+ return 0;
+ }
+
+ reuse_threshold = bitmap_popcount(bitmap_git.result) * REUSE_PERCENT;
+
+ if (reuse_objects < reuse_threshold)
+ return -1;
+
+ bitmap_git.reuse_objects = *entries = reuse_objects;
+ *up_to = bitmap_git.reverse_index->revindex[reuse_objects].offset;
+ *packfile = bitmap_git.pack;
+
+ return 0;
+}
+
+void traverse_bitmap_commit_list(show_reachable_fn show_reachable)
+{
+ assert(bitmap_git.result);
+
+ show_objects_for_type(bitmap_git.result, bitmap_git.commits,
+ OBJ_COMMIT, show_reachable);
+ show_objects_for_type(bitmap_git.result, bitmap_git.trees,
+ OBJ_TREE, show_reachable);
+ show_objects_for_type(bitmap_git.result, bitmap_git.blobs,
+ OBJ_BLOB, show_reachable);
+ show_objects_for_type(bitmap_git.result, bitmap_git.tags,
+ OBJ_TAG, show_reachable);
+
+ show_extended_objects(bitmap_git.result, show_reachable);
+
+ bitmap_free(bitmap_git.result);
+ bitmap_git.result = NULL;
+}
+
+static uint32_t count_object_type(struct bitmap *objects,
+ enum object_type type)
+{
+ struct eindex *eindex = &bitmap_git.ext_index;
+
+ uint32_t i = 0, count = 0;
+ struct ewah_iterator it;
+ eword_t filter;
+
+ switch (type) {
+ case OBJ_COMMIT:
+ ewah_iterator_init(&it, bitmap_git.commits);
+ break;
+
+ case OBJ_TREE:
+ ewah_iterator_init(&it, bitmap_git.trees);
+ break;
+
+ case OBJ_BLOB:
+ ewah_iterator_init(&it, bitmap_git.blobs);
+ break;
+
+ case OBJ_TAG:
+ ewah_iterator_init(&it, bitmap_git.tags);
+ break;
+
+ default:
+ return 0;
+ }
+
+ while (i < objects->word_alloc && ewah_iterator_next(&filter, &it)) {
+ eword_t word = objects->words[i++] & filter;
+ count += ewah_bit_popcount64(word);
+ }
+
+ for (i = 0; i < eindex->count; ++i) {
+ if (eindex->objects[i]->type == type &&
+ bitmap_get(objects, bitmap_git.pack->num_objects + i))
+ count++;
+ }
+
+ return count;
+}
+
+void count_bitmap_commit_list(uint32_t *commits, uint32_t *trees,
+ uint32_t *blobs, uint32_t *tags)
+{
+ assert(bitmap_git.result);
+
+ if (commits)
+ *commits = count_object_type(bitmap_git.result, OBJ_COMMIT);
+
+ if (trees)
+ *trees = count_object_type(bitmap_git.result, OBJ_TREE);
+
+ if (blobs)
+ *blobs = count_object_type(bitmap_git.result, OBJ_BLOB);
+
+ if (tags)
+ *tags = count_object_type(bitmap_git.result, OBJ_TAG);
+}
+
+struct bitmap_test_data {
+ struct bitmap *base;
+ struct progress *prg;
+ size_t seen;
+};
+
+static void test_show_object(struct object *object,
+ const struct name_path *path,
+ const char *last, void *data)
+{
+ struct bitmap_test_data *tdata = data;
+ int bitmap_pos;
+
+ bitmap_pos = bitmap_position(object->sha1);
+ if (bitmap_pos < 0)
+ die("Object not in bitmap: %s\n", sha1_to_hex(object->sha1));
+
+ bitmap_set(tdata->base, bitmap_pos);
+ display_progress(tdata->prg, ++tdata->seen);
+}
+
+static void test_show_commit(struct commit *commit, void *data)
+{
+ struct bitmap_test_data *tdata = data;
+ int bitmap_pos;
+
+ bitmap_pos = bitmap_position(commit->object.sha1);
+ if (bitmap_pos < 0)
+ die("Object not in bitmap: %s\n", sha1_to_hex(commit->object.sha1));
+
+ bitmap_set(tdata->base, bitmap_pos);
+ display_progress(tdata->prg, ++tdata->seen);
+}
+
+void test_bitmap_walk(struct rev_info *revs)
+{
+ struct object *root;
+ struct bitmap *result = NULL;
+ khiter_t pos;
+ size_t result_popcnt;
+ struct bitmap_test_data tdata;
+
+ if (prepare_bitmap_git())
+ die("failed to load bitmap indexes");
+
+ if (revs->pending.nr != 1)
+ die("you must specify exactly one commit to test");
+
+ fprintf(stderr, "Bitmap v%d test (%d entries loaded)\n",
+ bitmap_git.version, bitmap_git.entry_count);
+
+ root = revs->pending.objects[0].item;
+ pos = kh_get_sha1(bitmap_git.bitmaps, root->sha1);
+
+ if (pos < kh_end(bitmap_git.bitmaps)) {
+ struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, pos);
+ struct ewah_bitmap *bm = lookup_stored_bitmap(st);
+
+ fprintf(stderr, "Found bitmap for %s. %d bits / %08x checksum\n",
+ sha1_to_hex(root->sha1), (int)bm->bit_size, ewah_checksum(bm));
+
+ result = ewah_to_bitmap(bm);
+ }
+
+ if (result == NULL)
+ die("Commit %s doesn't have an indexed bitmap", sha1_to_hex(root->sha1));
+
+ revs->tag_objects = 1;
+ revs->tree_objects = 1;
+ revs->blob_objects = 1;
+
+ result_popcnt = bitmap_popcount(result);
+
+ if (prepare_revision_walk(revs))
+ die("revision walk setup failed");
+
+ tdata.base = bitmap_new();
+ tdata.prg = start_progress("Verifying bitmap entries", result_popcnt);
+ tdata.seen = 0;
+
+ traverse_commit_list(revs, &test_show_commit, &test_show_object, &tdata);
+
+ stop_progress(&tdata.prg);
+
+ if (bitmap_equals(result, tdata.base))
+ fprintf(stderr, "OK!\n");
+ else
+ fprintf(stderr, "Mismatch!\n");
+}
diff --git a/pack-bitmap.h b/pack-bitmap.h
new file mode 100644
index 0000000..9351aa5
--- /dev/null
+++ b/pack-bitmap.h
@@ -0,0 +1,43 @@
+#ifndef PACK_BITMAP_H
+#define PACK_BITMAP_H
+
+#include "ewah/ewok.h"
+#include "khash.h"
+
+struct bitmap_disk_entry {
+ uint32_t object_pos;
+ uint8_t xor_offset;
+ uint8_t flags;
+} __attribute__((packed));
+
+struct bitmap_disk_header {
+ char magic[4];
+ uint16_t version;
+ uint16_t options;
+ uint32_t entry_count;
+ unsigned char checksum[20];
+};
+
+static const char BITMAP_IDX_SIGNATURE[] = {'B', 'I', 'T', 'M'};;
+
+enum pack_bitmap_opts {
+ BITMAP_OPT_FULL_DAG = 1,
+};
+
+typedef int (*show_reachable_fn)(
+ const unsigned char *sha1,
+ enum object_type type,
+ int flags,
+ uint32_t hash,
+ struct packed_git *found_pack,
+ off_t found_offset);
+
+int prepare_bitmap_git(void);
+void count_bitmap_commit_list(uint32_t *commits, uint32_t *trees, uint32_t *blobs, uint32_t *tags);
+void traverse_bitmap_commit_list(show_reachable_fn show_reachable);
+void test_bitmap_walk(struct rev_info *revs);
+char *pack_bitmap_filename(struct packed_git *p);
+int prepare_bitmap_walk(struct rev_info *revs);
+int reuse_partial_packfile_from_bitmap(struct packed_git **packfile, uint32_t *entries, off_t *up_to);
+
+#endif
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 11/21] pack-objects: use bitmaps when packing objects
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (9 preceding siblings ...)
2013-11-14 12:44 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Jeff King
@ 2013-11-14 12:45 ` Jeff King
2013-12-07 15:47 ` Thomas Rast
2013-11-14 12:45 ` [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists Jeff King
` (10 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:45 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
In this patch, we use the bitmap API to perform the `Counting Objects`
phase in pack-objects, rather than a traditional walk through the object
graph. For a reasonably-packed large repo, the time to fetch and clone
is often dominated by the full-object revision walk during the Counting
Objects phase. Using bitmaps can reduce the CPU time required on the
server (and therefore start sending the actual pack data with less
delay).
For bitmaps to be used, the following must be true:
1. We must be packing to stdout (as a normal `pack-objects` from
`upload-pack` would do).
2. There must be a .bitmap index containing at least one of the
"have" objects that the client is asking for.
3. Bitmaps must be enabled (they are enabled by default, but can be
disabled by setting `pack.usebitmaps` to false, or by using
`--no-use-bitmap-index` on the command-line).
If any of these is not true, we fall back to doing a normal walk of the
object graph.
Here are some sample timings from a full pack of `torvalds/linux` (i.e.
something very similar to what would be generated for a clone of the
repository) that show the speedup produced by various
methods:
[existing graph traversal]
$ time git pack-objects --all --stdout --no-use-bitmap-index \
</dev/null >/dev/null
Counting objects: 3237103, done.
Compressing objects: 100% (508752/508752), done.
Total 3237103 (delta 2699584), reused 3237103 (delta 2699584)
real 0m44.111s
user 0m42.396s
sys 0m3.544s
[bitmaps only, without partial pack reuse; note that
pack reuse is automatic, so timing this required a
patch to disable it]
$ time git pack-objects --all --stdout </dev/null >/dev/null
Counting objects: 3237103, done.
Compressing objects: 100% (508752/508752), done.
Total 3237103 (delta 2699584), reused 3237103 (delta 2699584)
real 0m5.413s
user 0m5.604s
sys 0m1.804s
[bitmaps with pack reuse (what you get with this patch)]
$ time git pack-objects --all --stdout </dev/null >/dev/null
Reusing existing pack: 3237103, done.
Total 3237103 (delta 0), reused 0 (delta 0)
real 0m1.636s
user 0m1.460s
sys 0m0.172s
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/config.txt | 6 ++
builtin/pack-objects.c | 169 ++++++++++++++++++++++++++++++++++++++++-------
2 files changed, 150 insertions(+), 25 deletions(-)
diff --git a/Documentation/config.txt b/Documentation/config.txt
index ab26963..a981369 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -1858,6 +1858,12 @@ pack.packSizeLimit::
Common unit suffixes of 'k', 'm', or 'g' are
supported.
+pack.useBitmaps::
+ When true, git will use pack bitmaps (if available) when packing
+ to stdout (e.g., during the server side of a fetch). Defaults to
+ true. You should not generally need to turn this off unless
+ you are debugging pack bitmaps.
+
pager.<cmd>::
If the value is boolean, turns on or off pagination of the
output of a particular Git subcommand when writing to a tty.
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index faf746b..f04cce5 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -19,6 +19,7 @@
#include "refs.h"
#include "streaming.h"
#include "thread-utils.h"
+#include "pack-bitmap.h"
static const char *pack_usage[] = {
N_("git pack-objects --stdout [options...] [< ref-list | < object-list]"),
@@ -57,12 +58,23 @@ static struct progress *progress_state;
static int pack_compression_level = Z_DEFAULT_COMPRESSION;
static int pack_compression_seen;
+static struct packed_git *reuse_packfile;
+static uint32_t reuse_packfile_objects;
+static off_t reuse_packfile_offset;
+
+static int use_bitmap_index = 1;
+
static unsigned long delta_cache_size = 0;
static unsigned long max_delta_cache_size = 256 * 1024 * 1024;
static unsigned long cache_max_small_delta_size = 1000;
static unsigned long window_memory_limit = 0;
+enum {
+ OBJECT_ENTRY_EXCLUDE = (1 << 0),
+ OBJECT_ENTRY_NO_TRY_DELTA = (1 << 1)
+};
+
/*
* stats
*/
@@ -678,6 +690,49 @@ static struct object_entry **compute_write_order(void)
return wo;
}
+static off_t write_reused_pack(struct sha1file *f)
+{
+ uint8_t buffer[8192];
+ off_t to_write;
+ int fd;
+
+ if (!is_pack_valid(reuse_packfile))
+ return 0;
+
+ fd = git_open_noatime(reuse_packfile->pack_name);
+ if (fd < 0)
+ return 0;
+
+ if (lseek(fd, sizeof(struct pack_header), SEEK_SET) == -1) {
+ close(fd);
+ return 0;
+ }
+
+ if (reuse_packfile_offset < 0)
+ reuse_packfile_offset = reuse_packfile->pack_size - 20;
+
+ to_write = reuse_packfile_offset - sizeof(struct pack_header);
+
+ while (to_write) {
+ int read_pack = xread(fd, buffer, sizeof(buffer));
+
+ if (read_pack <= 0) {
+ close(fd);
+ return 0;
+ }
+
+ if (read_pack > to_write)
+ read_pack = to_write;
+
+ sha1write(f, buffer, read_pack);
+ to_write -= read_pack;
+ }
+
+ close(fd);
+ written += reuse_packfile_objects;
+ return reuse_packfile_offset - sizeof(struct pack_header);
+}
+
static void write_pack_file(void)
{
uint32_t i = 0, j;
@@ -704,6 +759,18 @@ static void write_pack_file(void)
offset = write_pack_header(f, nr_remaining);
if (!offset)
die_errno("unable to write pack header");
+
+ if (reuse_packfile) {
+ off_t packfile_size;
+ assert(pack_to_stdout);
+
+ packfile_size = write_reused_pack(f);
+ if (!packfile_size)
+ die_errno("failed to re-use existing pack");
+
+ offset += packfile_size;
+ }
+
nr_written = 0;
for (; i < to_pack.nr_objects; i++) {
struct object_entry *e = write_order[i];
@@ -800,14 +867,14 @@ static int no_try_delta(const char *path)
return 0;
}
-static int add_object_entry(const unsigned char *sha1, enum object_type type,
- const char *name, int exclude)
+static int add_object_entry_1(const unsigned char *sha1, enum object_type type,
+ int flags, uint32_t name_hash,
+ struct packed_git *found_pack, off_t found_offset)
{
struct object_entry *entry;
- struct packed_git *p, *found_pack = NULL;
- off_t found_offset = 0;
- uint32_t hash = pack_name_hash(name);
+ struct packed_git *p;
uint32_t index_pos;
+ int exclude = (flags & OBJECT_ENTRY_EXCLUDE);
entry = packlist_find(&to_pack, sha1, &index_pos);
if (entry) {
@@ -822,36 +889,42 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
if (!exclude && local && has_loose_object_nonlocal(sha1))
return 0;
- for (p = packed_git; p; p = p->next) {
- off_t offset = find_pack_entry_one(sha1, p);
- if (offset) {
- if (!found_pack) {
- if (!is_pack_valid(p)) {
- warning("packfile %s cannot be accessed", p->pack_name);
- continue;
+ if (!found_pack) {
+ for (p = packed_git; p; p = p->next) {
+ off_t offset = find_pack_entry_one(sha1, p);
+ if (offset) {
+ if (!found_pack) {
+ if (!is_pack_valid(p)) {
+ warning("packfile %s cannot be accessed", p->pack_name);
+ continue;
+ }
+ found_offset = offset;
+ found_pack = p;
}
- found_offset = offset;
- found_pack = p;
+ if (exclude)
+ break;
+ if (incremental)
+ return 0;
+ if (local && !p->pack_local)
+ return 0;
+ if (ignore_packed_keep && p->pack_local && p->pack_keep)
+ return 0;
}
- if (exclude)
- break;
- if (incremental)
- return 0;
- if (local && !p->pack_local)
- return 0;
- if (ignore_packed_keep && p->pack_local && p->pack_keep)
- return 0;
}
}
entry = packlist_alloc(&to_pack, sha1, index_pos);
- entry->hash = hash;
+ entry->hash = name_hash;
if (type)
entry->type = type;
if (exclude)
entry->preferred_base = 1;
else
nr_result++;
+
+ if (flags & OBJECT_ENTRY_NO_TRY_DELTA)
+ entry->no_try_delta = 1;
+
if (found_pack) {
entry->in_pack = found_pack;
entry->in_pack_offset = found_offset;
@@ -859,10 +932,21 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
display_progress(progress_state, to_pack.nr_objects);
+ return 1;
+}
+
+static int add_object_entry(const unsigned char *sha1, enum object_type type,
+ const char *name, int exclude)
+{
+ int flags = 0;
+
+ if (exclude)
+ flags |= OBJECT_ENTRY_EXCLUDE;
+
if (name && no_try_delta(name))
- entry->no_try_delta = 1;
+ flags |= OBJECT_ENTRY_NO_TRY_DELTA;
- return 1;
+ return add_object_entry_1(sha1, type, flags, pack_name_hash(name), NULL, 0);
}
struct pbase_tree_cache {
@@ -2027,6 +2111,10 @@ static int git_pack_config(const char *k, const char *v, void *cb)
cache_max_small_delta_size = git_config_int(k, v);
return 0;
}
+ if (!strcmp(k, "pack.usebitmaps")) {
+ use_bitmap_index = git_config_bool(k, v);
+ return 0;
+ }
if (!strcmp(k, "pack.threads")) {
delta_search_threads = git_config_int(k, v);
if (delta_search_threads < 0)
@@ -2235,6 +2323,29 @@ static void loosen_unused_packed_objects(struct rev_info *revs)
}
}
+static int get_object_list_from_bitmap(struct rev_info *revs)
+{
+ if (prepare_bitmap_walk(revs) < 0)
+ return -1;
+
+ if (!reuse_partial_packfile_from_bitmap(
+ &reuse_packfile,
+ &reuse_packfile_objects,
+ &reuse_packfile_offset)) {
+ assert(reuse_packfile_objects);
+ nr_result += reuse_packfile_objects;
+
+ if (progress) {
+ fprintf(stderr, "Reusing existing pack: %d, done.\n",
+ reuse_packfile_objects);
+ fflush(stderr);
+ }
+ }
+
+ traverse_bitmap_commit_list(&add_object_entry_1);
+ return 0;
+}
+
static void get_object_list(int ac, const char **av)
{
struct rev_info revs;
@@ -2262,6 +2373,9 @@ static void get_object_list(int ac, const char **av)
die("bad revision '%s'", line);
}
+ if (use_bitmap_index && !get_object_list_from_bitmap(&revs))
+ return;
+
if (prepare_revision_walk(&revs))
die("revision walk setup failed");
mark_edges_uninteresting(&revs, show_edge);
@@ -2391,6 +2505,8 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
N_("pack compression level")),
OPT_SET_INT(0, "keep-true-parents", &grafts_replace_parents,
N_("do not hide commits by grafts"), 0),
+ OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index,
+ N_("use a bitmap index if available to speed up counting objects")),
OPT_END(),
};
@@ -2457,6 +2573,9 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
if (keep_unreachable && unpack_unreachable)
die("--keep-unreachable and --unpack-unreachable are incompatible.");
+ if (!use_internal_rev_list || !pack_to_stdout || is_repository_shallow())
+ use_bitmap_index = 0;
+
if (progress && all_progress_implied)
progress = 2;
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (10 preceding siblings ...)
2013-11-14 12:45 ` [PATCH v3 11/21] pack-objects: use bitmaps when packing objects Jeff King
@ 2013-11-14 12:45 ` Jeff King
2013-12-07 16:05 ` Thomas Rast
2013-11-14 12:45 ` [PATCH v3 13/21] pack-objects: implement bitmap writing Jeff King
` (9 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:45 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
The bitmap reachability index used to speed up the counting objects
phase during `pack-objects` can also be used to optimize a normal
rev-list if the only thing required are the SHA1s of the objects during
the list (i.e., not the path names at which trees and blobs were found).
Calling `git rev-list --objects --use-bitmap-index [committish]` will
perform an object iteration based on a bitmap result instead of actually
walking the object graph.
These are some example timings for `torvalds/linux` (warm cache,
best-of-five):
$ time git rev-list --objects master > /dev/null
real 0m34.191s
user 0m33.904s
sys 0m0.268s
$ time git rev-list --objects --use-bitmap-index master > /dev/null
real 0m1.041s
user 0m0.976s
sys 0m0.064s
Likewise, using `git rev-list --count --use-bitmap-index` will speed up
the counting operation by building the resulting bitmap and performing a
fast popcount (number of bits set on the bitmap) on the result.
Here are some sample timings of different ways to count commits in
`torvalds/linux`:
$ time git rev-list master | wc -l
399882
real 0m6.524s
user 0m6.060s
sys 0m3.284s
$ time git rev-list --count master
399882
real 0m4.318s
user 0m4.236s
sys 0m0.076s
$ time git rev-list --use-bitmap-index --count master
399882
real 0m0.217s
user 0m0.176s
sys 0m0.040s
This also respects negative refs, so you can use it to count
a slice of history:
$ time git rev-list --count v3.0..master
144843
real 0m1.971s
user 0m1.932s
sys 0m0.036s
$ time git rev-list --use-bitmap-index --count v3.0..master
real 0m0.280s
user 0m0.220s
sys 0m0.056s
Though note that the closer the endpoints, the less it helps. In the
traversal case, we have fewer commits to cross, so we take less time.
But the bitmap time is dominated by generating the pack revindex, which
is constant with respect to the refs given.
Note that you cannot yet get a fast --left-right count of a symmetric
difference (e.g., "--count --left-right master...topic"). The slow part
of that walk actually happens during the merge-base determination when
we parse "master...topic". Even though a count does not actually need to
know the real merge base (it only needs to take the symmetric difference
of the bitmaps), the revision code would require some refactoring to
handle this case.
Additionally, a `--test-bitmap` flag has been added that will perform
the same rev-list manually (i.e. using a normal revwalk) and using
bitmaps, and verify that the results are the same. This can be used to
exercise the bitmap code, and also to verify that the contents of the
.bitmap file are sane.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/git-rev-list.txt | 1 +
Documentation/rev-list-options.txt | 8 ++++++++
builtin/rev-list.c | 39 ++++++++++++++++++++++++++++++++++++++
3 files changed, 48 insertions(+)
diff --git a/Documentation/git-rev-list.txt b/Documentation/git-rev-list.txt
index 045b37b..7a1585d 100644
--- a/Documentation/git-rev-list.txt
+++ b/Documentation/git-rev-list.txt
@@ -55,6 +55,7 @@ SYNOPSIS
[ \--reverse ]
[ \--walk-reflogs ]
[ \--no-walk ] [ \--do-walk ]
+ [ \--use-bitmap-index ]
<commit>... [ \-- <paths>... ]
DESCRIPTION
diff --git a/Documentation/rev-list-options.txt b/Documentation/rev-list-options.txt
index ec86d09..f7c8a4d 100644
--- a/Documentation/rev-list-options.txt
+++ b/Documentation/rev-list-options.txt
@@ -274,6 +274,14 @@ See also linkgit:git-reflog[1].
Output excluded boundary commits. Boundary commits are
prefixed with `-`.
+ifdef::git-rev-list[]
+--use-bitmap-index::
+
+ Try to speed up the traversal using the pack bitmap index (if
+ one is available). Note that when traversing with `--objects`,
+ trees and blobs will not have their associated path printed.
+endif::git-rev-list[]
+
--
History Simplification
diff --git a/builtin/rev-list.c b/builtin/rev-list.c
index 0745e2d..9f92905 100644
--- a/builtin/rev-list.c
+++ b/builtin/rev-list.c
@@ -3,6 +3,8 @@
#include "diff.h"
#include "revision.h"
#include "list-objects.h"
+#include "pack.h"
+#include "pack-bitmap.h"
#include "builtin.h"
#include "log-tree.h"
#include "graph.h"
@@ -257,6 +259,18 @@ static int show_bisect_vars(struct rev_list_info *info, int reaches, int all)
return 0;
}
+static int show_object_fast(
+ const unsigned char *sha1,
+ enum object_type type,
+ int exclude,
+ uint32_t name_hash,
+ struct packed_git *found_pack,
+ off_t found_offset)
+{
+ fprintf(stdout, "%s\n", sha1_to_hex(sha1));
+ return 1;
+}
+
int cmd_rev_list(int argc, const char **argv, const char *prefix)
{
struct rev_info revs;
@@ -265,6 +279,7 @@ int cmd_rev_list(int argc, const char **argv, const char *prefix)
int bisect_list = 0;
int bisect_show_vars = 0;
int bisect_find_all = 0;
+ int use_bitmap_index = 0;
git_config(git_default_config, NULL);
init_revisions(&revs, prefix);
@@ -306,6 +321,14 @@ int cmd_rev_list(int argc, const char **argv, const char *prefix)
bisect_show_vars = 1;
continue;
}
+ if (!strcmp(arg, "--use-bitmap-index")) {
+ use_bitmap_index = 1;
+ continue;
+ }
+ if (!strcmp(arg, "--test-bitmap")) {
+ test_bitmap_walk(&revs);
+ return 0;
+ }
usage(rev_list_usage);
}
@@ -333,6 +356,22 @@ int cmd_rev_list(int argc, const char **argv, const char *prefix)
if (bisect_list)
revs.limited = 1;
+ if (use_bitmap_index) {
+ if (revs.count && !revs.left_right && !revs.cherry_mark) {
+ uint32_t commit_count;
+ if (!prepare_bitmap_walk(&revs)) {
+ count_bitmap_commit_list(&commit_count, NULL, NULL, NULL);
+ printf("%d\n", commit_count);
+ return 0;
+ }
+ } else if (revs.tag_objects && revs.tree_objects && revs.blob_objects) {
+ if (!prepare_bitmap_walk(&revs)) {
+ traverse_bitmap_commit_list(&show_object_fast);
+ return 0;
+ }
+ }
+ }
+
if (prepare_revision_walk(&revs))
die("revision walk setup failed");
if (revs.tree_objects)
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 13/21] pack-objects: implement bitmap writing
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (11 preceding siblings ...)
2013-11-14 12:45 ` [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists Jeff King
@ 2013-11-14 12:45 ` Jeff King
2013-12-07 16:32 ` Thomas Rast
2013-11-14 12:45 ` [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts) Jeff King
` (8 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:45 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
This commit extends more the functionality of `pack-objects` by allowing
it to write out a `.bitmap` index next to any written packs, together
with the `.idx` index that currently gets written.
If bitmap writing is enabled for a given repository (either by calling
`pack-objects` with the `--write-bitmap-index` flag or by having
`pack.writebitmaps` set to `true` in the config) and pack-objects is
writing a packfile that would normally be indexed (i.e. not piping to
stdout), we will attempt to write the corresponding bitmap index for the
packfile.
Bitmap index writing happens after the packfile and its index has been
successfully written to disk (`finish_tmp_packfile`). The process is
performed in several steps:
1. `bitmap_writer_set_checksum`: this call stores the partial
checksum for the packfile being written; the checksum will be
written in the resulting bitmap index to verify its integrity
2. `bitmap_writer_build_type_index`: this call uses the array of
`struct object_entry` that has just been sorted when writing out
the actual packfile index to disk to generate 4 type-index bitmaps
(one for each object type).
These bitmaps have their nth bit set if the given object is of
the bitmap's type. E.g. the nth bit of the Commits bitmap will be
1 if the nth object in the packfile index is a commit.
This is a very cheap operation because the bitmap writing code has
access to the metadata stored in the `struct object_entry` array,
and hence the real type for each object in the packfile.
3. `bitmap_writer_reuse_bitmaps`: if there exists an existing bitmap
index for one of the packfiles we're trying to repack, this call
will efficiently rebuild the existing bitmaps so they can be
reused on the new index. All the existing bitmaps will be stored
in a `reuse` hash table, and the commit selection phase will
prioritize these when selecting, as they can be written directly
to the new index without having to perform a revision walk to
fill the bitmap. This can greatly speed up the repack of a
repository that already has bitmaps.
4. `bitmap_writer_select_commits`: if bitmap writing is enabled for
a given `pack-objects` run, the sequence of commits generated
during the Counting Objects phase will be stored in an array.
We then use that array to build up the list of selected commits.
Writing a bitmap in the index for each object in the repository
would be cost-prohibitive, so we use a simple heuristic to pick
the commits that will be indexed with bitmaps.
The current heuristics are a simplified version of JGit's
original implementation. We select a higher density of commits
depending on their age: the 100 most recent commits are always
selected, after that we pick 1 commit of each 100, and the gap
increases as the commits grow older. On top of that, we make sure
that every single branch that has not been merged (all the tips
that would be required from a clone) gets their own bitmap, and
when selecting commits between a gap, we tend to prioritize the
commit with the most parents.
Do note that there is no right/wrong way to perform commit
selection; different selection algorithms will result in
different commits being selected, but there's no such thing as
"missing a commit". The bitmap walker algorithm implemented in
`prepare_bitmap_walk` is able to adapt to missing bitmaps by
performing manual walks that complete the bitmap: the ideal
selection algorithm, however, would select the commits that are
more likely to be used as roots for a walk in the future (e.g.
the tips of each branch, and so on) to ensure a bitmap for them
is always available.
5. `bitmap_writer_build`: this is the computationally expensive part
of bitmap generation. Based on the list of commits that were
selected in the previous step, we perform several incremental
walks to generate the bitmap for each commit.
The walks begin from the oldest commit, and are built up
incrementally for each branch. E.g. consider this dag where A, B,
C, D, E, F are the selected commits, and a, b, c, e are a chunk
of simplified history that will not receive bitmaps.
A---a---B--b--C--c--D
\
E--e--F
We start by building the bitmap for A, using A as the root for a
revision walk and marking all the objects that are reachable
until the walk is over. Once this bitmap is stored, we reuse the
bitmap walker to perform the walk for B, assuming that once we
reach A again, the walk will be terminated because A has already
been SEEN on the previous walk.
This process is repeated for C, and D, but when we try to
generate the bitmaps for E, we can reuse neither the current walk
nor the bitmap we have generated so far.
What we do now is resetting both the walk and clearing the
bitmap, and performing the walk from scratch using E as the
origin. This new walk, however, does not need to be completed.
Once we hit B, we can lookup the bitmap we have already stored
for that commit and OR it with the existing bitmap we've composed
so far, allowing us to limit the walk early.
After all the bitmaps have been generated, another iteration
through the list of commits is performed to find the best XOR
offsets for compression before writing them to disk. Because of
the incremental nature of these bitmaps, XORing one of them with
its predecesor results in a minimal "bitmap delta" most of the
time. We can write this delta to the on-disk bitmap index, and
then re-compose the original bitmaps by XORing them again when
loaded.
This is a phase very similar to pack-object's `find_delta` (using
bitmaps instead of objects, of course), except the heuristics
have been greatly simplified: we only check the 10 bitmaps before
any given one to find best compressing one. This gives good
results in practice, because there is locality in the ordering of
the objects (and therefore bitmaps) in the packfile.
6. `bitmap_writer_finish`: the last step in the process is
serializing to disk all the bitmap data that has been generated
in the two previous steps.
The bitmap is written to a tmp file and then moved atomically to
its final destination, using the same process as
`pack-write.c:write_idx_file`.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/config.txt | 8 +
Makefile | 1 +
builtin/pack-objects.c | 53 +++++
pack-bitmap-write.c | 535 +++++++++++++++++++++++++++++++++++++++++++++++
pack-bitmap.c | 92 ++++++++
pack-bitmap.h | 19 ++
pack-objects.h | 1 +
pack-write.c | 2 +
8 files changed, 711 insertions(+)
create mode 100644 pack-bitmap-write.c
diff --git a/Documentation/config.txt b/Documentation/config.txt
index a981369..a439a4c 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -1864,6 +1864,14 @@ pack.useBitmaps::
true. You should not generally need to turn this off unless
you are debugging pack bitmaps.
+pack.writebitmaps::
+ When true, git will write a bitmap index when packing all
+ objects to disk (e.g., as when `git repack -a` is run). This
+ index can speed up the "counting objects" phase of subsequent
+ packs created for clones and fetches, at the cost of some disk
+ space and extra time spent on the initial repack. Defaults to
+ false.
+
pager.<cmd>::
If the value is boolean, turns on or off pagination of the
output of a particular Git subcommand when writing to a tty.
diff --git a/Makefile b/Makefile
index b983d78..555d44c 100644
--- a/Makefile
+++ b/Makefile
@@ -839,6 +839,7 @@ LIB_OBJS += notes-merge.o
LIB_OBJS += notes-utils.o
LIB_OBJS += object.o
LIB_OBJS += pack-bitmap.o
+LIB_OBJS += pack-bitmap-write.o
LIB_OBJS += pack-check.o
LIB_OBJS += pack-objects.o
LIB_OBJS += pack-revindex.o
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index f04cce5..26646e7 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -63,6 +63,7 @@ static uint32_t reuse_packfile_objects;
static off_t reuse_packfile_offset;
static int use_bitmap_index = 1;
+static int write_bitmap_index;
static unsigned long delta_cache_size = 0;
static unsigned long max_delta_cache_size = 256 * 1024 * 1024;
@@ -81,6 +82,24 @@ enum {
static uint32_t written, written_delta;
static uint32_t reused, reused_delta;
+/*
+ * Indexed commits
+ */
+static struct commit **indexed_commits;
+static unsigned int indexed_commits_nr;
+static unsigned int indexed_commits_alloc;
+
+static void index_commit_for_bitmap(struct commit *commit)
+{
+ if (indexed_commits_nr >= indexed_commits_alloc) {
+ indexed_commits_alloc = (indexed_commits_alloc + 32) * 2;
+ indexed_commits = xrealloc(indexed_commits,
+ indexed_commits_alloc * sizeof(struct commit *));
+ }
+
+ indexed_commits[indexed_commits_nr++] = commit;
+}
+
static void *get_delta(struct object_entry *entry)
{
unsigned long size, base_size, delta_size;
@@ -823,9 +842,30 @@ static void write_pack_file(void)
if (sizeof(tmpname) <= strlen(base_name) + 50)
die("pack base name '%s' too long", base_name);
snprintf(tmpname, sizeof(tmpname), "%s-", base_name);
+
+ if (write_bitmap_index) {
+ bitmap_writer_set_checksum(sha1);
+ bitmap_writer_build_type_index(written_list, nr_written);
+ }
+
finish_tmp_packfile(tmpname, pack_tmp_name,
written_list, nr_written,
&pack_idx_opts, sha1);
+
+ if (write_bitmap_index) {
+ char *end_of_name_prefix = strrchr(tmpname, 0);
+ sprintf(end_of_name_prefix, "%s.bitmap", sha1_to_hex(sha1));
+
+ stop_progress(&progress_state);
+
+ bitmap_writer_show_progress(progress);
+ bitmap_writer_reuse_bitmaps(&to_pack);
+ bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1);
+ bitmap_writer_build(&to_pack);
+ bitmap_writer_finish(written_list, nr_written, tmpname);
+ write_bitmap_index = 0;
+ }
+
free(pack_tmp_name);
puts(sha1_to_hex(sha1));
}
@@ -2111,6 +2151,10 @@ static int git_pack_config(const char *k, const char *v, void *cb)
cache_max_small_delta_size = git_config_int(k, v);
return 0;
}
+ if (!strcmp(k, "pack.writebitmaps")) {
+ write_bitmap_index = git_config_bool(k, v);
+ return 0;
+ }
if (!strcmp(k, "pack.usebitmaps")) {
use_bitmap_index = git_config_bool(k, v);
return 0;
@@ -2173,6 +2217,9 @@ static void show_commit(struct commit *commit, void *data)
{
add_object_entry(commit->object.sha1, OBJ_COMMIT, NULL, 0);
commit->object.flags |= OBJECT_ADDED;
+
+ if (write_bitmap_index)
+ index_commit_for_bitmap(commit);
}
static void show_object(struct object *obj,
@@ -2365,6 +2412,7 @@ static void get_object_list(int ac, const char **av)
if (*line == '-') {
if (!strcmp(line, "--not")) {
flags ^= UNINTERESTING;
+ write_bitmap_index = 0;
continue;
}
die("not a rev '%s'", line);
@@ -2507,6 +2555,8 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
N_("do not hide commits by grafts"), 0),
OPT_BOOL(0, "use-bitmap-index", &use_bitmap_index,
N_("use a bitmap index if available to speed up counting objects")),
+ OPT_BOOL(0, "write-bitmap-index", &write_bitmap_index,
+ N_("write a bitmap index together with the pack index")),
OPT_END(),
};
@@ -2576,6 +2626,9 @@ int cmd_pack_objects(int argc, const char **argv, const char *prefix)
if (!use_internal_rev_list || !pack_to_stdout || is_repository_shallow())
use_bitmap_index = 0;
+ if (pack_to_stdout || !rev_list_all)
+ write_bitmap_index = 0;
+
if (progress && all_progress_implied)
progress = 2;
diff --git a/pack-bitmap-write.c b/pack-bitmap-write.c
new file mode 100644
index 0000000..954a74d
--- /dev/null
+++ b/pack-bitmap-write.c
@@ -0,0 +1,535 @@
+#include "cache.h"
+#include "commit.h"
+#include "tag.h"
+#include "diff.h"
+#include "revision.h"
+#include "list-objects.h"
+#include "progress.h"
+#include "pack-revindex.h"
+#include "pack.h"
+#include "pack-bitmap.h"
+#include "sha1-lookup.h"
+#include "pack-objects.h"
+
+struct bitmapped_commit {
+ struct commit *commit;
+ struct ewah_bitmap *bitmap;
+ struct ewah_bitmap *write_as;
+ int flags;
+ int xor_offset;
+ uint32_t commit_pos;
+};
+
+struct bitmap_writer {
+ struct ewah_bitmap *commits;
+ struct ewah_bitmap *trees;
+ struct ewah_bitmap *blobs;
+ struct ewah_bitmap *tags;
+
+ khash_sha1 *bitmaps;
+ khash_sha1 *reused;
+ struct packing_data *to_pack;
+
+ struct bitmapped_commit *selected;
+ unsigned int selected_nr, selected_alloc;
+
+ struct progress *progress;
+ int show_progress;
+ unsigned char pack_checksum[20];
+};
+
+static struct bitmap_writer writer;
+
+void bitmap_writer_show_progress(int show)
+{
+ writer.show_progress = show;
+}
+
+/**
+ * Build the initial type index for the packfile
+ */
+void bitmap_writer_build_type_index(struct pack_idx_entry **index,
+ uint32_t index_nr)
+{
+ uint32_t i;
+
+ writer.commits = ewah_new();
+ writer.trees = ewah_new();
+ writer.blobs = ewah_new();
+ writer.tags = ewah_new();
+
+ for (i = 0; i < index_nr; ++i) {
+ struct object_entry *entry = (struct object_entry *)index[i];
+ enum object_type real_type;
+
+ entry->in_pack_pos = i;
+
+ switch (entry->type) {
+ case OBJ_COMMIT:
+ case OBJ_TREE:
+ case OBJ_BLOB:
+ case OBJ_TAG:
+ real_type = entry->type;
+ break;
+
+ default:
+ real_type = sha1_object_info(entry->idx.sha1, NULL);
+ break;
+ }
+
+ switch (real_type) {
+ case OBJ_COMMIT:
+ ewah_set(writer.commits, i);
+ break;
+
+ case OBJ_TREE:
+ ewah_set(writer.trees, i);
+ break;
+
+ case OBJ_BLOB:
+ ewah_set(writer.blobs, i);
+ break;
+
+ case OBJ_TAG:
+ ewah_set(writer.tags, i);
+ break;
+
+ default:
+ die("Missing type information for %s (%d/%d)",
+ sha1_to_hex(entry->idx.sha1), real_type, entry->type);
+ }
+ }
+}
+
+/**
+ * Compute the actual bitmaps
+ */
+static struct object **seen_objects;
+static unsigned int seen_objects_nr, seen_objects_alloc;
+
+static inline void push_bitmapped_commit(struct commit *commit, struct ewah_bitmap *reused)
+{
+ if (writer.selected_nr >= writer.selected_alloc) {
+ writer.selected_alloc = (writer.selected_alloc + 32) * 2;
+ writer.selected = xrealloc(writer.selected,
+ writer.selected_alloc * sizeof(struct bitmapped_commit));
+ }
+
+ writer.selected[writer.selected_nr].commit = commit;
+ writer.selected[writer.selected_nr].bitmap = reused;
+ writer.selected[writer.selected_nr].flags = 0;
+
+ writer.selected_nr++;
+}
+
+static inline void mark_as_seen(struct object *object)
+{
+ ALLOC_GROW(seen_objects, seen_objects_nr + 1, seen_objects_alloc);
+ seen_objects[seen_objects_nr++] = object;
+}
+
+static inline void reset_all_seen(void)
+{
+ unsigned int i;
+ for (i = 0; i < seen_objects_nr; ++i) {
+ seen_objects[i]->flags &= ~(SEEN | ADDED | SHOWN);
+ }
+ seen_objects_nr = 0;
+}
+
+static uint32_t find_object_pos(const unsigned char *sha1)
+{
+ struct object_entry *entry = packlist_find(writer.to_pack, sha1, NULL);
+
+ if (!entry) {
+ die("Failed to write bitmap index. Packfile doesn't have full closure "
+ "(object %s is missing)", sha1_to_hex(sha1));
+ }
+
+ return entry->in_pack_pos;
+}
+
+static void show_object(struct object *object, const struct name_path *path,
+ const char *last, void *data)
+{
+ struct bitmap *base = data;
+ bitmap_set(base, find_object_pos(object->sha1));
+ mark_as_seen(object);
+}
+
+static void show_commit(struct commit *commit, void *data)
+{
+ mark_as_seen((struct object *)commit);
+}
+
+static int
+add_to_include_set(struct bitmap *base, struct commit *commit)
+{
+ khiter_t hash_pos;
+ uint32_t bitmap_pos = find_object_pos(commit->object.sha1);
+
+ if (bitmap_get(base, bitmap_pos))
+ return 0;
+
+ hash_pos = kh_get_sha1(writer.bitmaps, commit->object.sha1);
+ if (hash_pos < kh_end(writer.bitmaps)) {
+ struct bitmapped_commit *bc = kh_value(writer.bitmaps, hash_pos);
+ bitmap_or_ewah(base, bc->bitmap);
+ return 0;
+ }
+
+ bitmap_set(base, bitmap_pos);
+ return 1;
+}
+
+static int
+should_include(struct commit *commit, void *_data)
+{
+ struct bitmap *base = _data;
+
+ if (!add_to_include_set(base, commit)) {
+ struct commit_list *parent = commit->parents;
+
+ mark_as_seen((struct object *)commit);
+
+ while (parent) {
+ parent->item->object.flags |= SEEN;
+ mark_as_seen((struct object *)parent->item);
+ parent = parent->next;
+ }
+
+ return 0;
+ }
+
+ return 1;
+}
+
+static void compute_xor_offsets(void)
+{
+ static const int MAX_XOR_OFFSET_SEARCH = 10;
+
+ int i, next = 0;
+
+ while (next < writer.selected_nr) {
+ struct bitmapped_commit *stored = &writer.selected[next];
+
+ int best_offset = 0;
+ struct ewah_bitmap *best_bitmap = stored->bitmap;
+ struct ewah_bitmap *test_xor;
+
+ for (i = 1; i <= MAX_XOR_OFFSET_SEARCH; ++i) {
+ int curr = next - i;
+
+ if (curr < 0)
+ break;
+
+ test_xor = ewah_pool_new();
+ ewah_xor(writer.selected[curr].bitmap, stored->bitmap, test_xor);
+
+ if (test_xor->buffer_size < best_bitmap->buffer_size) {
+ if (best_bitmap != stored->bitmap)
+ ewah_pool_free(best_bitmap);
+
+ best_bitmap = test_xor;
+ best_offset = i;
+ } else {
+ ewah_pool_free(test_xor);
+ }
+ }
+
+ stored->xor_offset = best_offset;
+ stored->write_as = best_bitmap;
+
+ next++;
+ }
+}
+
+void bitmap_writer_build(struct packing_data *to_pack)
+{
+ static const double REUSE_BITMAP_THRESHOLD = 0.2;
+
+ int i, reuse_after, need_reset;
+ struct bitmap *base = bitmap_new();
+ struct rev_info revs;
+
+ writer.bitmaps = kh_init_sha1();
+ writer.to_pack = to_pack;
+
+ if (writer.show_progress)
+ writer.progress = start_progress("Building bitmaps", writer.selected_nr);
+
+ init_revisions(&revs, NULL);
+ revs.tag_objects = 1;
+ revs.tree_objects = 1;
+ revs.blob_objects = 1;
+ revs.no_walk = 0;
+
+ revs.include_check = should_include;
+ reset_revision_walk();
+
+ reuse_after = writer.selected_nr * REUSE_BITMAP_THRESHOLD;
+ need_reset = 0;
+
+ for (i = writer.selected_nr - 1; i >= 0; --i) {
+ struct bitmapped_commit *stored;
+ struct object *object;
+
+ khiter_t hash_pos;
+ int hash_ret;
+
+ stored = &writer.selected[i];
+ object = (struct object *)stored->commit;
+
+ if (stored->bitmap == NULL) {
+ if (i < writer.selected_nr - 1 &&
+ (need_reset ||
+ !in_merge_bases(writer.selected[i + 1].commit,
+ stored->commit))) {
+ bitmap_reset(base);
+ reset_all_seen();
+ }
+
+ add_pending_object(&revs, object, "");
+ revs.include_check_data = base;
+
+ if (prepare_revision_walk(&revs))
+ die("revision walk setup failed");
+
+ traverse_commit_list(&revs, show_commit, show_object, base);
+
+ revs.pending.nr = 0;
+ revs.pending.alloc = 0;
+ revs.pending.objects = NULL;
+
+ stored->bitmap = bitmap_to_ewah(base);
+ need_reset = 0;
+ } else
+ need_reset = 1;
+
+ if (i >= reuse_after)
+ stored->flags |= BITMAP_FLAG_REUSE;
+
+ hash_pos = kh_put_sha1(writer.bitmaps, object->sha1, &hash_ret);
+ if (hash_ret == 0)
+ die("Duplicate entry when writing index: %s",
+ sha1_to_hex(object->sha1));
+
+ kh_value(writer.bitmaps, hash_pos) = stored;
+ display_progress(writer.progress, writer.selected_nr - i);
+ }
+
+ bitmap_free(base);
+ stop_progress(&writer.progress);
+
+ compute_xor_offsets();
+}
+
+/**
+ * Select the commits that will be bitmapped
+ */
+static inline unsigned int next_commit_index(unsigned int idx)
+{
+ static const unsigned int MIN_COMMITS = 100;
+ static const unsigned int MAX_COMMITS = 5000;
+
+ static const unsigned int MUST_REGION = 100;
+ static const unsigned int MIN_REGION = 20000;
+
+ unsigned int offset, next;
+
+ if (idx <= MUST_REGION)
+ return 0;
+
+ if (idx <= MIN_REGION) {
+ offset = idx - MUST_REGION;
+ return (offset < MIN_COMMITS) ? offset : MIN_COMMITS;
+ }
+
+ offset = idx - MIN_REGION;
+ next = (offset < MAX_COMMITS) ? offset : MAX_COMMITS;
+
+ return (next > MIN_COMMITS) ? next : MIN_COMMITS;
+}
+
+static int date_compare(const void *_a, const void *_b)
+{
+ struct commit *a = *(struct commit **)_a;
+ struct commit *b = *(struct commit **)_b;
+ return (long)b->date - (long)a->date;
+}
+
+void bitmap_writer_reuse_bitmaps(struct packing_data *to_pack)
+{
+ if (prepare_bitmap_git() < 0)
+ return;
+
+ writer.reused = kh_init_sha1();
+ rebuild_existing_bitmaps(to_pack, writer.reused, writer.show_progress);
+}
+
+static struct ewah_bitmap *find_reused_bitmap(const unsigned char *sha1)
+{
+ khiter_t hash_pos;
+
+ if (!writer.reused)
+ return NULL;
+
+ hash_pos = kh_get_sha1(writer.reused, sha1);
+ if (hash_pos >= kh_end(writer.reused))
+ return NULL;
+
+ return kh_value(writer.reused, hash_pos);
+}
+
+void bitmap_writer_select_commits(struct commit **indexed_commits,
+ unsigned int indexed_commits_nr,
+ int max_bitmaps)
+{
+ unsigned int i = 0, j, next;
+
+ qsort(indexed_commits, indexed_commits_nr, sizeof(indexed_commits[0]),
+ date_compare);
+
+ if (writer.show_progress)
+ writer.progress = start_progress("Selecting bitmap commits", 0);
+
+ if (indexed_commits_nr < 100) {
+ for (i = 0; i < indexed_commits_nr; ++i)
+ push_bitmapped_commit(indexed_commits[i], NULL);
+ return;
+ }
+
+ for (;;) {
+ struct ewah_bitmap *reused_bitmap = NULL;
+ struct commit *chosen = NULL;
+
+ next = next_commit_index(i);
+
+ if (i + next >= indexed_commits_nr)
+ break;
+
+ if (max_bitmaps > 0 && writer.selected_nr >= max_bitmaps) {
+ writer.selected_nr = max_bitmaps;
+ break;
+ }
+
+ if (next == 0) {
+ chosen = indexed_commits[i];
+ reused_bitmap = find_reused_bitmap(chosen->object.sha1);
+ } else {
+ chosen = indexed_commits[i + next];
+
+ for (j = 0; j <= next; ++j) {
+ struct commit *cm = indexed_commits[i + j];
+
+ reused_bitmap = find_reused_bitmap(cm->object.sha1);
+ if (reused_bitmap || (cm->object.flags & NEEDS_BITMAP) != 0) {
+ chosen = cm;
+ break;
+ }
+
+ if (cm->parents && cm->parents->next)
+ chosen = cm;
+ }
+ }
+
+ push_bitmapped_commit(chosen, reused_bitmap);
+
+ i += next + 1;
+ display_progress(writer.progress, i);
+ }
+
+ stop_progress(&writer.progress);
+}
+
+
+static int sha1write_ewah_helper(void *f, const void *buf, size_t len)
+{
+ /* sha1write will die on error */
+ sha1write(f, buf, len);
+ return len;
+}
+
+/**
+ * Write the bitmap index to disk
+ */
+static inline void dump_bitmap(struct sha1file *f, struct ewah_bitmap *bitmap)
+{
+ if (ewah_serialize_to(bitmap, sha1write_ewah_helper, f) < 0)
+ die("Failed to write bitmap index");
+}
+
+static const unsigned char *sha1_access(size_t pos, void *table)
+{
+ struct pack_idx_entry **index = table;
+ return index[pos]->sha1;
+}
+
+static void write_selected_commits_v1(struct sha1file *f,
+ struct pack_idx_entry **index,
+ uint32_t index_nr)
+{
+ int i;
+
+ for (i = 0; i < writer.selected_nr; ++i) {
+ struct bitmapped_commit *stored = &writer.selected[i];
+ struct bitmap_disk_entry on_disk;
+
+ int commit_pos =
+ sha1_pos(stored->commit->object.sha1, index, index_nr, sha1_access);
+
+ if (commit_pos < 0)
+ die("BUG: trying to write commit not in index");
+
+ on_disk.object_pos = htonl(commit_pos);
+ on_disk.xor_offset = stored->xor_offset;
+ on_disk.flags = stored->flags;
+
+ sha1write(f, &on_disk, sizeof(on_disk));
+ dump_bitmap(f, stored->write_as);
+ }
+}
+
+void bitmap_writer_set_checksum(unsigned char *sha1)
+{
+ hashcpy(writer.pack_checksum, sha1);
+}
+
+void bitmap_writer_finish(struct pack_idx_entry **index,
+ uint32_t index_nr,
+ const char *filename)
+{
+ static char tmp_file[PATH_MAX];
+ static uint16_t default_version = 1;
+ static uint16_t flags = BITMAP_OPT_FULL_DAG;
+ struct sha1file *f;
+
+ struct bitmap_disk_header header;
+
+ int fd = odb_mkstemp(tmp_file, sizeof(tmp_file), "pack/tmp_bitmap_XXXXXX");
+
+ if (fd < 0)
+ die_errno("unable to create '%s'", tmp_file);
+ f = sha1fd(fd, tmp_file);
+
+ memcpy(header.magic, BITMAP_IDX_SIGNATURE, sizeof(BITMAP_IDX_SIGNATURE));
+ header.version = htons(default_version);
+ header.options = htons(flags);
+ header.entry_count = htonl(writer.selected_nr);
+ memcpy(header.checksum, writer.pack_checksum, 20);
+
+ sha1write(f, &header, sizeof(header));
+ dump_bitmap(f, writer.commits);
+ dump_bitmap(f, writer.trees);
+ dump_bitmap(f, writer.blobs);
+ dump_bitmap(f, writer.tags);
+ write_selected_commits_v1(f, index, index_nr);
+
+ sha1close(f, NULL, CSUM_FSYNC);
+
+ if (adjust_shared_perm(tmp_file))
+ die_errno("unable to make temporary bitmap file readable");
+
+ if (rename(tmp_file, filename))
+ die_errno("unable to rename temporary bitmap file to '%s'", filename);
+}
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 340f02b..7792dfd 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -968,3 +968,95 @@ void test_bitmap_walk(struct rev_info *revs)
else
fprintf(stderr, "Mismatch!\n");
}
+
+static int rebuild_bitmap(uint32_t *reposition,
+ struct ewah_bitmap *source,
+ struct bitmap *dest)
+{
+ uint32_t pos = 0;
+ struct ewah_iterator it;
+ eword_t word;
+
+ ewah_iterator_init(&it, source);
+
+ while (ewah_iterator_next(&word, &it)) {
+ uint32_t offset, bit_pos;
+
+ for (offset = 0; offset < BITS_IN_WORD; ++offset) {
+ if ((word >> offset) == 0)
+ break;
+
+ offset += ewah_bit_ctz64(word >> offset);
+
+ bit_pos = reposition[pos + offset];
+ if (bit_pos > 0)
+ bitmap_set(dest, bit_pos - 1);
+ else /* can't reuse, we don't have the object */
+ return -1;
+ }
+
+ pos += BITS_IN_WORD;
+ }
+ return 0;
+}
+
+int rebuild_existing_bitmaps(struct packing_data *mapping,
+ khash_sha1 *reused_bitmaps,
+ int show_progress)
+{
+ uint32_t i, num_objects;
+ uint32_t *reposition;
+ struct bitmap *rebuild;
+ struct stored_bitmap *stored;
+ struct progress *progress = NULL;
+
+ khiter_t hash_pos;
+ int hash_ret;
+
+ if (prepare_bitmap_git() < 0)
+ return -1;
+
+ num_objects = bitmap_git.pack->num_objects;
+ reposition = xcalloc(num_objects, sizeof(uint32_t));
+
+ for (i = 0; i < num_objects; ++i) {
+ const unsigned char *sha1;
+ struct revindex_entry *entry;
+ struct object_entry *oe;
+
+ entry = &bitmap_git.reverse_index->revindex[i];
+ sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
+ oe = packlist_find(mapping, sha1, NULL);
+
+ if (oe)
+ reposition[i] = oe->in_pack_pos + 1;
+ }
+
+ rebuild = bitmap_new();
+ i = 0;
+
+ if (show_progress)
+ progress = start_progress("Reusing bitmaps", 0);
+
+ kh_foreach_value(bitmap_git.bitmaps, stored, {
+ if (stored->flags & BITMAP_FLAG_REUSE) {
+ if (!rebuild_bitmap(reposition,
+ lookup_stored_bitmap(stored),
+ rebuild)) {
+ hash_pos = kh_put_sha1(reused_bitmaps,
+ stored->sha1,
+ &hash_ret);
+ kh_value(reused_bitmaps, hash_pos) =
+ bitmap_to_ewah(rebuild);
+ }
+ bitmap_reset(rebuild);
+ display_progress(progress, ++i);
+ }
+ });
+
+ stop_progress(&progress);
+
+ free(reposition);
+ bitmap_free(rebuild);
+ return 0;
+}
diff --git a/pack-bitmap.h b/pack-bitmap.h
index 9351aa5..5e3fb9a 100644
--- a/pack-bitmap.h
+++ b/pack-bitmap.h
@@ -3,6 +3,7 @@
#include "ewah/ewok.h"
#include "khash.h"
+#include "pack-objects.h"
struct bitmap_disk_entry {
uint32_t object_pos;
@@ -20,10 +21,16 @@ struct bitmap_disk_header {
static const char BITMAP_IDX_SIGNATURE[] = {'B', 'I', 'T', 'M'};;
+#define NEEDS_BITMAP (1u<<22)
+
enum pack_bitmap_opts {
BITMAP_OPT_FULL_DAG = 1,
};
+enum pack_bitmap_flags {
+ BITMAP_FLAG_REUSE = 0x1
+};
+
typedef int (*show_reachable_fn)(
const unsigned char *sha1,
enum object_type type,
@@ -39,5 +46,17 @@ void test_bitmap_walk(struct rev_info *revs);
char *pack_bitmap_filename(struct packed_git *p);
int prepare_bitmap_walk(struct rev_info *revs);
int reuse_partial_packfile_from_bitmap(struct packed_git **packfile, uint32_t *entries, off_t *up_to);
+int rebuild_existing_bitmaps(struct packing_data *mapping, khash_sha1 *reused_bitmaps, int show_progress);
+
+void bitmap_writer_show_progress(int show);
+void bitmap_writer_set_checksum(unsigned char *sha1);
+void bitmap_writer_build_type_index(struct pack_idx_entry **index, uint32_t index_nr);
+void bitmap_writer_reuse_bitmaps(struct packing_data *to_pack);
+void bitmap_writer_select_commits(struct commit **indexed_commits,
+ unsigned int indexed_commits_nr, int max_bitmaps);
+void bitmap_writer_build(struct packing_data *to_pack);
+void bitmap_writer_finish(struct pack_idx_entry **index,
+ uint32_t index_nr,
+ const char *filename);
#endif
diff --git a/pack-objects.h b/pack-objects.h
index 90ad0a8..d1b98b3 100644
--- a/pack-objects.h
+++ b/pack-objects.h
@@ -17,6 +17,7 @@ struct object_entry {
enum object_type type;
enum object_type in_pack_type; /* could be delta */
uint32_t hash; /* name hint hash */
+ unsigned int in_pack_pos;
unsigned char in_pack_header_size;
unsigned preferred_base:1; /*
* we do not pack this, but is available
diff --git a/pack-write.c b/pack-write.c
index ca9e63b..6203d37 100644
--- a/pack-write.c
+++ b/pack-write.c
@@ -371,5 +371,7 @@ void finish_tmp_packfile(char *name_buffer,
if (rename(idx_tmp_name, name_buffer))
die_errno("unable to rename temporary index file");
+ *end_of_name_prefix = '\0';
+
free((void *)idx_tmp_name);
}
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts)
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (12 preceding siblings ...)
2013-11-14 12:45 ` [PATCH v3 13/21] pack-objects: implement bitmap writing Jeff King
@ 2013-11-14 12:45 ` Jeff King
2013-12-07 16:34 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 15/21] repack: turn exts array into array-of-struct Jeff King
` (7 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:45 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
We have a static array of extensions, but hardcode the size
of the array in our loops. Let's pull out this magic number,
which will make it easier to change.
Signed-off-by: Jeff King <peff@peff.net>
---
builtin/repack.c | 8 ++++----
1 file changed, 4 insertions(+), 4 deletions(-)
diff --git a/builtin/repack.c b/builtin/repack.c
index a0ff5c7..2e88975 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -115,7 +115,7 @@ static void remove_redundant_pack(const char *dir_name, const char *base_name)
int cmd_repack(int argc, const char **argv, const char *prefix)
{
- const char *exts[2] = {".pack", ".idx"};
+ const char *exts[] = {".pack", ".idx"};
struct child_process cmd;
struct string_list_item *item;
struct argv_array cmd_args = ARGV_ARRAY_INIT;
@@ -258,7 +258,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
*/
failed = 0;
for_each_string_list_item(item, &names) {
- for (ext = 0; ext < 2; ext++) {
+ for (ext = 0; ext < ARRAY_SIZE(exts); ext++) {
char *fname, *fname_old;
fname = mkpathdup("%s/%s%s", packdir,
item->string, exts[ext]);
@@ -315,7 +315,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
/* Now the ones with the same name are out of the way... */
for_each_string_list_item(item, &names) {
- for (ext = 0; ext < 2; ext++) {
+ for (ext = 0; ext < ARRAY_SIZE(exts); ext++) {
char *fname, *fname_old;
struct stat statbuffer;
fname = mkpathdup("%s/pack-%s%s",
@@ -335,7 +335,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
/* Remove the "old-" files */
for_each_string_list_item(item, &names) {
- for (ext = 0; ext < 2; ext++) {
+ for (ext = 0; ext < ARRAY_SIZE(exts); ext++) {
char *fname;
fname = mkpath("%s/old-pack-%s%s",
packdir,
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 15/21] repack: turn exts array into array-of-struct
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (13 preceding siblings ...)
2013-11-14 12:45 ` [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts) Jeff King
@ 2013-11-14 12:46 ` Jeff King
2013-12-07 16:34 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 16/21] repack: handle optional files created by pack-objects Jeff King
` (6 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:46 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
This is slightly more verbose, but will let us annotate the
extensions with further options in future commits.
Signed-off-by: Jeff King <peff@peff.net>
---
builtin/repack.c | 17 +++++++++++------
1 file changed, 11 insertions(+), 6 deletions(-)
diff --git a/builtin/repack.c b/builtin/repack.c
index 2e88975..a176de2 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -115,7 +115,12 @@ static void remove_redundant_pack(const char *dir_name, const char *base_name)
int cmd_repack(int argc, const char **argv, const char *prefix)
{
- const char *exts[] = {".pack", ".idx"};
+ struct {
+ const char *name;
+ } exts[] = {
+ {".pack"},
+ {".idx"},
+ };
struct child_process cmd;
struct string_list_item *item;
struct argv_array cmd_args = ARGV_ARRAY_INIT;
@@ -261,14 +266,14 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
for (ext = 0; ext < ARRAY_SIZE(exts); ext++) {
char *fname, *fname_old;
fname = mkpathdup("%s/%s%s", packdir,
- item->string, exts[ext]);
+ item->string, exts[ext].name);
if (!file_exists(fname)) {
free(fname);
continue;
}
fname_old = mkpath("%s/old-%s%s", packdir,
- item->string, exts[ext]);
+ item->string, exts[ext].name);
if (file_exists(fname_old))
if (unlink(fname_old))
failed = 1;
@@ -319,9 +324,9 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
char *fname, *fname_old;
struct stat statbuffer;
fname = mkpathdup("%s/pack-%s%s",
- packdir, item->string, exts[ext]);
+ packdir, item->string, exts[ext].name);
fname_old = mkpathdup("%s-%s%s",
- packtmp, item->string, exts[ext]);
+ packtmp, item->string, exts[ext].name);
if (!stat(fname_old, &statbuffer)) {
statbuffer.st_mode &= ~(S_IWUSR | S_IWGRP | S_IWOTH);
chmod(fname_old, statbuffer.st_mode);
@@ -340,7 +345,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
fname = mkpath("%s/old-pack-%s%s",
packdir,
item->string,
- exts[ext]);
+ exts[ext].name);
if (remove_path(fname))
warning(_("removing '%s' failed"), fname);
}
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 16/21] repack: handle optional files created by pack-objects
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (14 preceding siblings ...)
2013-11-14 12:46 ` [PATCH v3 15/21] repack: turn exts array into array-of-struct Jeff King
@ 2013-11-14 12:46 ` Jeff King
2013-12-07 16:35 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 17/21] repack: consider bitmaps when performing repacks Jeff King
` (5 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:46 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
We ask pack-objects to pack to a set of temporary files, and
then rename them into place. Some files that pack-objects
creates may be optional (like a .bitmap file), in which case
we would not want to call rename(). We already call stat()
and make the chmod optional if the file cannot be accessed.
We could simply skip the rename step in this case, but that
would be a minor regression in noticing problems with
non-optional files (like the .pack and .idx files).
Instead, we can now annotate extensions as optional, and
skip them if they don't exist (and otherwise rely on
rename() to barf).
Signed-off-by: Jeff King <peff@peff.net>
---
builtin/repack.c | 9 +++++++--
1 file changed, 7 insertions(+), 2 deletions(-)
diff --git a/builtin/repack.c b/builtin/repack.c
index a176de2..8b7dfd0 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -117,6 +117,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
{
struct {
const char *name;
+ unsigned optional:1;
} exts[] = {
{".pack"},
{".idx"},
@@ -323,6 +324,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
for (ext = 0; ext < ARRAY_SIZE(exts); ext++) {
char *fname, *fname_old;
struct stat statbuffer;
+ int exists = 0;
fname = mkpathdup("%s/pack-%s%s",
packdir, item->string, exts[ext].name);
fname_old = mkpathdup("%s-%s%s",
@@ -330,9 +332,12 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
if (!stat(fname_old, &statbuffer)) {
statbuffer.st_mode &= ~(S_IWUSR | S_IWGRP | S_IWOTH);
chmod(fname_old, statbuffer.st_mode);
+ exists = 1;
+ }
+ if (exists || !exts[ext].optional) {
+ if (rename(fname_old, fname))
+ die_errno(_("renaming '%s' failed"), fname_old);
}
- if (rename(fname_old, fname))
- die_errno(_("renaming '%s' failed"), fname_old);
free(fname);
free(fname_old);
}
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 17/21] repack: consider bitmaps when performing repacks
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (15 preceding siblings ...)
2013-11-14 12:46 ` [PATCH v3 16/21] repack: handle optional files created by pack-objects Jeff King
@ 2013-11-14 12:46 ` Jeff King
2013-12-07 16:37 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking Jeff King
` (4 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:46 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
Since `pack-objects` will write a `.bitmap` file next to the `.pack` and
`.idx` files, this commit teaches `git-repack` to consider the new
bitmap indexes (if they exist) when performing repack operations.
This implies moving old bitmap indexes out of the way if we are
repacking a repository that already has them, and moving the newly
generated bitmap indexes into the `objects/pack` directory, next to
their corresponding packfiles.
Since `git repack` is now capable of handling these `.bitmap` files,
a normal `git gc` run on a repository that has `pack.writebitmaps` set
to true in its config file will generate bitmap indexes as part of the
garbage collection process.
Alternatively, `git repack` can be called with the `-b` switch to
explicitly generate bitmap indexes if you are experimenting
and don't want them on all the time.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/git-repack.txt | 9 ++++++++-
builtin/repack.c | 9 ++++++++-
2 files changed, 16 insertions(+), 2 deletions(-)
diff --git a/Documentation/git-repack.txt b/Documentation/git-repack.txt
index 509cf73..002cfd5 100644
--- a/Documentation/git-repack.txt
+++ b/Documentation/git-repack.txt
@@ -9,7 +9,7 @@ git-repack - Pack unpacked objects in a repository
SYNOPSIS
--------
[verse]
-'git repack' [-a] [-A] [-d] [-f] [-F] [-l] [-n] [-q] [--window=<n>] [--depth=<n>]
+'git repack' [-a] [-A] [-d] [-f] [-F] [-l] [-n] [-q] [-b] [--window=<n>] [--depth=<n>]
DESCRIPTION
-----------
@@ -110,6 +110,13 @@ other objects in that pack they already have locally.
The default is unlimited, unless the config variable
`pack.packSizeLimit` is set.
+-b::
+--write-bitmap-index::
+ Write a reachability bitmap index as part of the repack. This
+ only makes sense when used with `-a` or `-A`, as the bitmaps
+ must be able to refer to all reachable objects. This option
+ overrides the setting of `pack.writebitmaps`.
+
Configuration
-------------
diff --git a/builtin/repack.c b/builtin/repack.c
index 8b7dfd0..239f278 100644
--- a/builtin/repack.c
+++ b/builtin/repack.c
@@ -94,7 +94,7 @@ static void get_non_kept_pack_filenames(struct string_list *fname_list)
static void remove_redundant_pack(const char *dir_name, const char *base_name)
{
- const char *exts[] = {".pack", ".idx", ".keep"};
+ const char *exts[] = {".pack", ".idx", ".keep", ".bitmap"};
int i;
struct strbuf buf = STRBUF_INIT;
size_t plen;
@@ -121,6 +121,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
} exts[] = {
{".pack"},
{".idx"},
+ {".bitmap", 1},
};
struct child_process cmd;
struct string_list_item *item;
@@ -143,6 +144,7 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
int no_update_server_info = 0;
int quiet = 0;
int local = 0;
+ int write_bitmap = -1;
struct option builtin_repack_options[] = {
OPT_BIT('a', NULL, &pack_everything,
@@ -161,6 +163,8 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
OPT__QUIET(&quiet, N_("be quiet")),
OPT_BOOL('l', "local", &local,
N_("pass --local to git-pack-objects")),
+ OPT_BOOL('b', "write-bitmap-index", &write_bitmap,
+ N_("write bitmap index")),
OPT_STRING(0, "unpack-unreachable", &unpack_unreachable, N_("approxidate"),
N_("with -A, do not loosen objects older than this")),
OPT_INTEGER(0, "window", &window,
@@ -202,6 +206,9 @@ int cmd_repack(int argc, const char **argv, const char *prefix)
argv_array_pushf(&cmd_args, "--no-reuse-delta");
if (no_reuse_object)
argv_array_pushf(&cmd_args, "--no-reuse-object");
+ if (write_bitmap >= 0)
+ argv_array_pushf(&cmd_args, "--%swrite-bitmap-index",
+ write_bitmap ? "" : "no-");
if (pack_everything & ALL_INTO_ONE) {
get_non_kept_pack_filenames(&existing_packs);
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (16 preceding siblings ...)
2013-11-14 12:46 ` [PATCH v3 17/21] repack: consider bitmaps when performing repacks Jeff King
@ 2013-11-14 12:46 ` Jeff King
2013-12-07 16:38 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 19/21] t: add basic bitmap functionality tests Jeff King
` (3 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:46 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Count-objects will report any "garbage" files in the packs
directory, including files whose extensions it does not
know (case 1), and files whose matching ".pack" file is
missing (case 2). Without having learned about ".bitmap"
files, the current code reports all such files as garbage
(case 1), even if their pack exists. Instead, they should be
treated as case 2.
Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
sha1_file.c | 1 +
1 file changed, 1 insertion(+)
diff --git a/sha1_file.c b/sha1_file.c
index 5557bd9..d810b58 100644
--- a/sha1_file.c
+++ b/sha1_file.c
@@ -1194,6 +1194,7 @@ static void prepare_packed_git_one(char *objdir, int local)
if (has_extension(de->d_name, ".idx") ||
has_extension(de->d_name, ".pack") ||
+ has_extension(de->d_name, ".bitmap") ||
has_extension(de->d_name, ".keep"))
string_list_append(&garbage, path);
else
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 19/21] t: add basic bitmap functionality tests
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (17 preceding siblings ...)
2013-11-14 12:46 ` [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking Jeff King
@ 2013-11-14 12:46 ` Jeff King
2013-12-07 16:43 ` Thomas Rast
2013-11-14 12:48 ` [PATCH v3 20/21] t/perf: add tests for pack bitmaps Jeff King
` (2 subsequent siblings)
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:46 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
Now that we can read and write bitmaps, we can exercise them
with some basic functionality tests. These tests aren't
particularly useful for seeing the benefit, as the test
repo is too small for it to make a difference. However, we
can at least check that using bitmaps does not break anything.
Signed-off-by: Jeff King <peff@peff.net>
---
t/t5310-pack-bitmaps.sh | 138 ++++++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 138 insertions(+)
create mode 100755 t/t5310-pack-bitmaps.sh
diff --git a/t/t5310-pack-bitmaps.sh b/t/t5310-pack-bitmaps.sh
new file mode 100755
index 0000000..bd743f4
--- /dev/null
+++ b/t/t5310-pack-bitmaps.sh
@@ -0,0 +1,138 @@
+#!/bin/sh
+
+test_description='exercise basic bitmap functionality'
+. ./test-lib.sh
+
+test_expect_success 'setup repo with moderate-sized history' '
+ for i in $(test_seq 1 10); do
+ test_commit $i
+ done &&
+ git checkout -b other HEAD~5 &&
+ for i in $(test_seq 1 10); do
+ test_commit side-$i
+ done &&
+ git checkout master &&
+ blob=$(echo tagged-blob | git hash-object -w --stdin) &&
+ git tag tagged-blob $blob &&
+ git config pack.writebitmaps true
+'
+
+test_expect_success 'full repack creates bitmaps' '
+ git repack -ad &&
+ ls .git/objects/pack/ | grep bitmap >output &&
+ test_line_count = 1 output
+'
+
+test_expect_success 'rev-list --test-bitmap verifies bitmaps' '
+ git rev-list --test-bitmap HEAD
+'
+
+rev_list_tests() {
+ state=$1
+
+ test_expect_success "counting commits via bitmap ($state)" '
+ git rev-list --count HEAD >expect &&
+ git rev-list --use-bitmap-index --count HEAD >actual &&
+ test_cmp expect actual
+ '
+
+ test_expect_success "counting partial commits via bitmap ($state)" '
+ git rev-list --count HEAD~5..HEAD >expect &&
+ git rev-list --use-bitmap-index --count HEAD~5..HEAD >actual &&
+ test_cmp expect actual
+ '
+
+ test_expect_success "counting non-linear history ($state)" '
+ git rev-list --count other...master >expect &&
+ git rev-list --use-bitmap-index --count other...master >actual &&
+ test_cmp expect actual
+ '
+
+ test_expect_success "enumerate --objects ($state)" '
+ git rev-list --objects --use-bitmap-index HEAD >tmp &&
+ cut -d" " -f1 <tmp >tmp2 &&
+ sort <tmp2 >actual &&
+ git rev-list --objects HEAD >tmp &&
+ cut -d" " -f1 <tmp >tmp2 &&
+ sort <tmp2 >expect &&
+ test_cmp expect actual
+ '
+
+ test_expect_success "bitmap --objects handles non-commit objects ($state)" '
+ git rev-list --objects --use-bitmap-index HEAD tagged-blob >actual &&
+ grep $blob actual
+ '
+}
+
+rev_list_tests 'full bitmap'
+
+test_expect_success 'clone from bitmapped repository' '
+ git clone --no-local --bare . clone.git &&
+ git rev-parse HEAD >expect &&
+ git --git-dir=clone.git rev-parse HEAD >actual &&
+ test_cmp expect actual
+'
+
+test_expect_success 'setup further non-bitmapped commits' '
+ for i in $(test_seq 1 10); do
+ test_commit further-$i
+ done
+'
+
+rev_list_tests 'partial bitmap'
+
+test_expect_success 'fetch (partial bitmap)' '
+ git --git-dir=clone.git fetch origin master:master &&
+ git rev-parse HEAD >expect &&
+ git --git-dir=clone.git rev-parse HEAD >actual &&
+ test_cmp expect actual
+'
+
+test_expect_success 'incremental repack cannot create bitmaps' '
+ test_commit more-1 &&
+ test_must_fail git repack -d
+'
+
+test_expect_success 'incremental repack can disable bitmaps' '
+ test_commit more-2 &&
+ git repack -d --no-write-bitmap-index
+'
+
+test_expect_success 'full repack, reusing previous bitmaps' '
+ git repack -ad &&
+ ls .git/objects/pack/ | grep bitmap >output &&
+ test_line_count = 1 output
+'
+
+test_expect_success 'fetch (full bitmap)' '
+ git --git-dir=clone.git fetch origin master:master &&
+ git rev-parse HEAD >expect &&
+ git --git-dir=clone.git rev-parse HEAD >actual &&
+ test_cmp expect actual
+'
+
+test_lazy_prereq JGIT '
+ type jgit
+'
+
+test_expect_success JGIT 'we can read jgit bitmaps' '
+ git clone . compat-jgit &&
+ (
+ cd compat-jgit &&
+ rm -f .git/objects/pack/*.bitmap &&
+ jgit gc &&
+ git rev-list --test-bitmap HEAD
+ )
+'
+
+test_expect_success JGIT 'jgit can read our bitmaps' '
+ git clone . compat-us.git &&
+ (
+ cd compat-us.git &&
+ git repack -adb &&
+ # jgit gc will barf if it does not like our bitmaps
+ jgit gc
+ )
+'
+
+test_done
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 20/21] t/perf: add tests for pack bitmaps
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (18 preceding siblings ...)
2013-11-14 12:46 ` [PATCH v3 19/21] t: add basic bitmap functionality tests Jeff King
@ 2013-11-14 12:48 ` Jeff King
2013-12-07 16:51 ` Thomas Rast
2013-11-14 12:48 ` [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache Jeff King
2013-11-14 19:19 ` [PATCH v3 0/21] pack bitmaps Ramsay Jones
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:48 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
This adds a few basic perf tests for the pack bitmap code to
show off its improvements. The tests are:
1. How long does it take to do a repack (it gets slower
with bitmaps, since we have to do extra work)?
2. How long does it take to do a clone (it gets faster
with bitmaps)?
3. How does a small fetch perform when we've just
repacked?
4. How does a clone perform when we haven't repacked since
a week of pushes?
Here are results against linux.git:
Test origin/master this tree
-----------------------------------------------------------------------
5310.2: repack to disk 33.64(32.64+2.04) 67.67(66.75+1.84) +101.2%
5310.3: simulated clone 30.49(29.47+2.05) 1.20(1.10+0.10) -96.1%
5310.4: simulated fetch 3.49(6.79+0.06) 5.57(22.35+0.07) +59.6%
5310.6: partial bitmap 36.70(43.87+1.81) 8.18(21.92+0.73) -77.7%
You can see that we do take longer to repack, but we do way
better for further clones. A small fetch performs a bit
worse, as we spend way more time on delta compression (note
the heavy user CPU time, as we have 8 threads) due to the
lack of name hashes for the bitmapped objects.
The final test shows how the bitmaps degrade over time
between packs. There's still a significant speedup over the
non-bitmap case, but we don't do quite as well (we have to
spend time accessing the "new" objects the old fashioned
way, including delta compression).
Signed-off-by: Jeff King <peff@peff.net>
---
t/perf/p5310-pack-bitmaps.sh | 56 ++++++++++++++++++++++++++++++++++++++++++++
1 file changed, 56 insertions(+)
create mode 100755 t/perf/p5310-pack-bitmaps.sh
diff --git a/t/perf/p5310-pack-bitmaps.sh b/t/perf/p5310-pack-bitmaps.sh
new file mode 100755
index 0000000..ac036d0
--- /dev/null
+++ b/t/perf/p5310-pack-bitmaps.sh
@@ -0,0 +1,56 @@
+#!/bin/sh
+
+test_description='Tests pack performance using bitmaps'
+. ./perf-lib.sh
+
+test_perf_large_repo
+
+# note that we do everything through config,
+# since we want to be able to compare bitmap-aware
+# git versus non-bitmap git
+test_expect_success 'setup bitmap config' '
+ git config pack.writebitmaps true
+'
+
+test_perf 'repack to disk' '
+ git repack -ad
+'
+
+test_perf 'simulated clone' '
+ git pack-objects --stdout --all </dev/null >/dev/null
+'
+
+test_perf 'simulated fetch' '
+ have=$(git rev-list HEAD --until=1.week.ago -1) &&
+ {
+ echo HEAD &&
+ echo ^$have
+ } | git pack-objects --revs --stdout >/dev/null
+'
+
+test_expect_success 'create partial bitmap state' '
+ # pick a commit to represent the repo tip in the past
+ cutoff=$(git rev-list HEAD --until=1.week.ago -1) &&
+ orig_tip=$(git rev-parse HEAD) &&
+
+ # now kill off all of the refs and pretend we had
+ # just the one tip
+ rm -rf .git/logs .git/refs/* .git/packed-refs
+ git update-ref HEAD $cutoff
+
+ # and then repack, which will leave us with a nice
+ # big bitmap pack of the "old" history, and all of
+ # the new history will be loose, as if it had been pushed
+ # up incrementally and exploded via unpack-objects
+ git repack -Ad
+
+ # and now restore our original tip, as if the pushes
+ # had happened
+ git update-ref HEAD $orig_tip
+'
+
+test_perf 'partial bitmap' '
+ git pack-objects --stdout --all </dev/null >/dev/null
+'
+
+test_done
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (19 preceding siblings ...)
2013-11-14 12:48 ` [PATCH v3 20/21] t/perf: add tests for pack bitmaps Jeff King
@ 2013-11-14 12:48 ` Jeff King
2013-12-07 16:59 ` Thomas Rast
2013-11-14 19:19 ` [PATCH v3 0/21] pack bitmaps Ramsay Jones
21 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-14 12:48 UTC (permalink / raw)
To: git; +Cc: Vicent Martí
From: Vicent Marti <tanoku@gmail.com>
When we use pack bitmaps rather than walking the object
graph, we end up with the list of objects to include in the
packfile, but we do not know the path at which any tree or
blob objects would be found.
In a recently packed repository, this is fine. A fetch would
use the paths only as a heuristic in the delta compression
phase, and a fully packed repository should not need to do
much delta compression.
As time passes, though, we may acquire more objects on top
of our large bitmapped pack. If clients fetch frequently,
then they never even look at the bitmapped history, and all
works as usual. However, a client who has not fetched since
the last bitmap repack will have "have" tips in the
bitmapped history, but "want" newer objects.
The bitmaps themselves degrade gracefully in this
circumstance. We manually walk the more recent bits of
history, and then use bitmaps when we hit them.
But we would also like to perform delta compression between
the newer objects and the bitmapped objects (both to delta
against what we know the user already has, but also between
"new" and "old" objects that the user is fetching). The lack
of pathnames makes our delta heuristics much less effective.
This patch adds an optional cache of the 32-bit name_hash
values to the end of the bitmap file. If present, a reader
can use it to match bitmapped and non-bitmapped names during
delta compression.
Here are perf results for p5310:
Test origin/master HEAD^ HEAD
-------------------------------------------------------------------------------------------------
5310.2: repack to disk 36.81(37.82+1.43) 47.70(48.74+1.41) +29.6% 47.75(48.70+1.51) +29.7%
5310.3: simulated clone 30.78(29.70+2.14) 1.08(0.97+0.10) -96.5% 1.07(0.94+0.12) -96.5%
5310.4: simulated fetch 3.16(6.10+0.08) 3.54(10.65+0.06) +12.0% 1.70(3.07+0.06) -46.2%
5310.6: partial bitmap 36.76(43.19+1.81) 6.71(11.25+0.76) -81.7% 4.08(6.26+0.46) -88.9%
You can see that the time spent on an incremental fetch goes
down, as our delta heuristics are able to do their work.
And we save time on the partial bitmap clone for the same
reason.
Signed-off-by: Vicent Marti <tanoku@gmail.com>
Signed-off-by: Jeff King <peff@peff.net>
---
Documentation/config.txt | 11 +++++++++++
Documentation/technical/bitmap-format.txt | 33 +++++++++++++++++++++++++++++++
builtin/pack-objects.c | 10 +++++++++-
pack-bitmap-write.c | 21 ++++++++++++++++++--
pack-bitmap.c | 11 +++++++++++
pack-bitmap.h | 4 +++-
t/perf/p5310-pack-bitmaps.sh | 3 ++-
t/t5310-pack-bitmaps.sh | 3 ++-
8 files changed, 90 insertions(+), 6 deletions(-)
diff --git a/Documentation/config.txt b/Documentation/config.txt
index a439a4c..e6d3922 100644
--- a/Documentation/config.txt
+++ b/Documentation/config.txt
@@ -1872,6 +1872,17 @@ pack.writebitmaps::
space and extra time spent on the initial repack. Defaults to
false.
+pack.writeBitmapHashCache::
+ When true, git will include a "hash cache" section in the bitmap
+ index (if one is written). This cache can be used to feed git's
+ delta heuristics, potentially leading to better deltas between
+ bitmapped and non-bitmapped objects (e.g., when serving a fetch
+ between an older, bitmapped pack and objects that have been
+ pushed since the last gc). The downside is that it consumes 4
+ bytes per object of disk space, and that JGit's bitmap
+ implementation does not understand it, causing it to complain if
+ Git and JGit are used on the same repository. Defaults to false.
+
pager.<cmd>::
If the value is boolean, turns on or off pagination of the
output of a particular Git subcommand when writing to a tty.
diff --git a/Documentation/technical/bitmap-format.txt b/Documentation/technical/bitmap-format.txt
index 7a86bd7..f8c18a0 100644
--- a/Documentation/technical/bitmap-format.txt
+++ b/Documentation/technical/bitmap-format.txt
@@ -21,6 +21,12 @@ GIT bitmap v1 format
requirement for the bitmap index format, also present in JGit,
that greatly reduces the complexity of the implementation.
+ - BITMAP_OPT_HASH_CACHE (0x4)
+ If present, the end of the bitmap file contains
+ `N` 32-bit name-hash values, one per object in the
+ pack. The format and meaning of the name-hash is
+ described below.
+
4-byte entry count (network byte order)
The total count of entries (bitmapped commits) in this bitmap index.
@@ -129,3 +135,30 @@ The bitstream represented by the above chunk is then:
The next word after `L_M` (if any) must again be a RLW, for the next
chunk. For efficient appending to the bitstream, the EWAH stores a
pointer to the last RLW in the stream.
+
+
+== Appendix B: Optional Bitmap Sections
+
+These sections may or may not be present in the `.bitmap` file; their
+presence is indicated by the header flags section described above.
+
+Name-hash cache
+---------------
+
+If the BITMAP_OPT_HASH_CACHE flag is set, the end of the bitmap contains
+a cache of 32-bit values, one per object in the pack. The value at
+position `i` is the hash of the pathname at which the `i`th object
+(counting in index order) in the pack can be found. This can be fed
+into the delta heuristics to compare objects with similar pathnames.
+
+The hash algorithm used is:
+
+ hash = 0;
+ while ((c = *name++))
+ if (!isspace(c))
+ hash = (hash >> 2) + (c << 24);
+
+Note that this hashing scheme is tied to the BITMAP_OPT_HASH_CACHE flag.
+If implementations want to choose a different hashing scheme, they are
+free to do so, but MUST allocate a new header flag (because comparing
+hashes made under two different schemes would be pointless).
diff --git a/builtin/pack-objects.c b/builtin/pack-objects.c
index 26646e7..4504789 100644
--- a/builtin/pack-objects.c
+++ b/builtin/pack-objects.c
@@ -64,6 +64,7 @@ static off_t reuse_packfile_offset;
static int use_bitmap_index = 1;
static int write_bitmap_index;
+static uint16_t write_bitmap_options;
static unsigned long delta_cache_size = 0;
static unsigned long max_delta_cache_size = 256 * 1024 * 1024;
@@ -862,7 +863,8 @@ static void write_pack_file(void)
bitmap_writer_reuse_bitmaps(&to_pack);
bitmap_writer_select_commits(indexed_commits, indexed_commits_nr, -1);
bitmap_writer_build(&to_pack);
- bitmap_writer_finish(written_list, nr_written, tmpname);
+ bitmap_writer_finish(written_list, nr_written,
+ tmpname, write_bitmap_options);
write_bitmap_index = 0;
}
@@ -2155,6 +2157,12 @@ static int git_pack_config(const char *k, const char *v, void *cb)
write_bitmap_index = git_config_bool(k, v);
return 0;
}
+ if (!strcmp(k, "pack.writebitmaphashcache")) {
+ if (git_config_bool(k, v))
+ write_bitmap_options |= BITMAP_OPT_HASH_CACHE;
+ else
+ write_bitmap_options &= ~BITMAP_OPT_HASH_CACHE;
+ }
if (!strcmp(k, "pack.usebitmaps")) {
use_bitmap_index = git_config_bool(k, v);
return 0;
diff --git a/pack-bitmap-write.c b/pack-bitmap-write.c
index 954a74d..1218bef 100644
--- a/pack-bitmap-write.c
+++ b/pack-bitmap-write.c
@@ -490,6 +490,19 @@ static void write_selected_commits_v1(struct sha1file *f,
}
}
+static void write_hash_cache(struct sha1file *f,
+ struct pack_idx_entry **index,
+ uint32_t index_nr)
+{
+ uint32_t i;
+
+ for (i = 0; i < index_nr; ++i) {
+ struct object_entry *entry = (struct object_entry *)index[i];
+ uint32_t hash_value = htonl(entry->hash);
+ sha1write(f, &hash_value, sizeof(hash_value));
+ }
+}
+
void bitmap_writer_set_checksum(unsigned char *sha1)
{
hashcpy(writer.pack_checksum, sha1);
@@ -497,7 +510,8 @@ void bitmap_writer_set_checksum(unsigned char *sha1)
void bitmap_writer_finish(struct pack_idx_entry **index,
uint32_t index_nr,
- const char *filename)
+ const char *filename,
+ uint16_t options)
{
static char tmp_file[PATH_MAX];
static uint16_t default_version = 1;
@@ -514,7 +528,7 @@ void bitmap_writer_finish(struct pack_idx_entry **index,
memcpy(header.magic, BITMAP_IDX_SIGNATURE, sizeof(BITMAP_IDX_SIGNATURE));
header.version = htons(default_version);
- header.options = htons(flags);
+ header.options = htons(flags | options);
header.entry_count = htonl(writer.selected_nr);
memcpy(header.checksum, writer.pack_checksum, 20);
@@ -525,6 +539,9 @@ void bitmap_writer_finish(struct pack_idx_entry **index,
dump_bitmap(f, writer.tags);
write_selected_commits_v1(f, index, index_nr);
+ if (options & BITMAP_OPT_HASH_CACHE)
+ write_hash_cache(f, index, index_nr);
+
sha1close(f, NULL, CSUM_FSYNC);
if (adjust_shared_perm(tmp_file))
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 7792dfd..078f7c6 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -66,6 +66,9 @@ static struct bitmap_index {
/* Number of bitmapped commits */
uint32_t entry_count;
+ /* Name-hash cache (or NULL if not present). */
+ uint32_t *hashes;
+
/*
* Extended index.
*
@@ -152,6 +155,11 @@ static int load_bitmap_header(struct bitmap_index *index)
if ((flags & BITMAP_OPT_FULL_DAG) == 0)
return error("Unsupported options for bitmap index file "
"(Git requires BITMAP_OPT_FULL_DAG)");
+
+ if (flags & BITMAP_OPT_HASH_CACHE) {
+ unsigned char *end = index->map + index->map_size - 20;
+ index->hashes = ((uint32_t *)end) - index->pack->num_objects;
+ }
}
index->entry_count = ntohl(header->entry_count);
@@ -626,6 +634,9 @@ static void show_objects_for_type(
entry = &bitmap_git.reverse_index->revindex[pos + offset];
sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
+ if (bitmap_git.hashes)
+ hash = ntohl(bitmap_git.hashes[entry->nr]);
+
show_reach(sha1, object_type, 0, hash, bitmap_git.pack, entry->offset);
}
diff --git a/pack-bitmap.h b/pack-bitmap.h
index 5e3fb9a..e4e1a57 100644
--- a/pack-bitmap.h
+++ b/pack-bitmap.h
@@ -25,6 +25,7 @@ static const char BITMAP_IDX_SIGNATURE[] = {'B', 'I', 'T', 'M'};;
enum pack_bitmap_opts {
BITMAP_OPT_FULL_DAG = 1,
+ BITMAP_OPT_HASH_CACHE = 4,
};
enum pack_bitmap_flags {
@@ -57,6 +58,7 @@ void bitmap_writer_select_commits(struct commit **indexed_commits,
void bitmap_writer_build(struct packing_data *to_pack);
void bitmap_writer_finish(struct pack_idx_entry **index,
uint32_t index_nr,
- const char *filename);
+ const char *filename,
+ uint16_t options);
#endif
diff --git a/t/perf/p5310-pack-bitmaps.sh b/t/perf/p5310-pack-bitmaps.sh
index ac036d0..8811fc4 100755
--- a/t/perf/p5310-pack-bitmaps.sh
+++ b/t/perf/p5310-pack-bitmaps.sh
@@ -9,7 +9,8 @@ test_perf_large_repo
# since we want to be able to compare bitmap-aware
# git versus non-bitmap git
test_expect_success 'setup bitmap config' '
- git config pack.writebitmaps true
+ git config pack.writebitmaps true &&
+ git config pack.writebitmaphashcache true
'
test_perf 'repack to disk' '
diff --git a/t/t5310-pack-bitmaps.sh b/t/t5310-pack-bitmaps.sh
index bd743f4..2c2632f 100755
--- a/t/t5310-pack-bitmaps.sh
+++ b/t/t5310-pack-bitmaps.sh
@@ -14,7 +14,8 @@ test_expect_success 'setup repo with moderate-sized history' '
git checkout master &&
blob=$(echo tagged-blob | git hash-object -w --stdin) &&
git tag tagged-blob $blob &&
- git config pack.writebitmaps true
+ git config pack.writebitmaps true &&
+ git config pack.writebitmaphashcache true
'
test_expect_success 'full repack creates bitmaps' '
--
1.8.5.rc0.443.g2df7f3f
^ permalink raw reply related [flat|nested] 55+ messages in thread
* Re: [PATCH v3 0/21] pack bitmaps
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
` (20 preceding siblings ...)
2013-11-14 12:48 ` [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache Jeff King
@ 2013-11-14 19:19 ` Ramsay Jones
2013-11-14 21:33 ` Jeff King
21 siblings, 1 reply; 55+ messages in thread
From: Ramsay Jones @ 2013-11-14 19:19 UTC (permalink / raw)
To: Jeff King, git; +Cc: Vicent Martí
On 14/11/13 12:41, Jeff King wrote:
> Here's another iteration of the pack bitmaps series. Compared to v2, it
> changes:
>
> - misc style/typo fixes
>
> - portability fixes from Ramsay and Torsten
Unfortunately, I didn't find time this weekend to finish the msvc build
fixes. However, after a quick squint at these patches, I think you have
almost done it for me! :-D
I must have misunderstood the previous discussion, because my patch was
written on the assumption that the ewah directory wouldn't be "git-ified"
(e.g. #include git-compat-util.h).
So, most of my patch is no longer necessary, given the use of the git
compat header (and removal of system headers). I suspect that you only
need to add an '#define PRIx64 "I64x"' definition (Hmm, probably to the
compat/mingw.h header).
I won't know for sure until I actually try them out, of course. I will
wait until these patches land in pu.
[Note: the msvc build is still broken, but the failure is not caused by
these patches. Unfortunately, the tests in t5310-*.sh fail. However, if
I include some debug code, the tests pass ... :-P ]
The part of the patch I was still working on was ...
>
> - count-objects garbage-reporting patch from Duy
>
> - disable bitmaps when is_repository_shallow(); this also covers the
> case where the client is shallow, since we feed pack-objects a
> --shallow-file in that case. This used to done by checking
> !internal_rev_list, but that doesn't apply after cdab485.
>
> - ewah sources now properly use git-compat-util.h and do not include
> system headers
>
> - the ewah code uses ewah_malloc, ewah_realloc, and so forth to let the
> project use a particular allocator (and we want to use xmalloc and
> friends). And we defined those in pack-bitmap.h, but of course that
> had no effect on the ewah/*.c files that did not include
> pack-bitmap.h. Since we are hacking up and git-ifying libewok
> anyway, we can just set the hardcoded fallback to xmalloc instead of
> malloc.
>
> - the ewah code used gcc's __builtin_ctzll, but did not provide a
> suitable fallback. We now provide a fallback in C.
... here.
I was messing around with several implementations (including the use of
msvc compiler intrinsics) with the intention of doing some timing tests
etc. [I suspected my C fallback function (a different implementation to
yours) would be slightly faster.]
ATB,
Ramsay Jones
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 0/21] pack bitmaps
2013-11-14 19:19 ` [PATCH v3 0/21] pack bitmaps Ramsay Jones
@ 2013-11-14 21:33 ` Jeff King
2013-11-14 23:09 ` Ramsay Jones
2013-11-16 10:28 ` Thomas Rast
0 siblings, 2 replies; 55+ messages in thread
From: Jeff King @ 2013-11-14 21:33 UTC (permalink / raw)
To: Ramsay Jones; +Cc: git, Vicent Martí
On Thu, Nov 14, 2013 at 07:19:38PM +0000, Ramsay Jones wrote:
> Unfortunately, I didn't find time this weekend to finish the msvc build
> fixes. However, after a quick squint at these patches, I think you have
> almost done it for me! :-D
>
> I must have misunderstood the previous discussion, because my patch was
> written on the assumption that the ewah directory wouldn't be "git-ified"
> (e.g. #include git-compat-util.h).
I think it was up for debate at some point, but we did decide to go
ahead and git-ify. Please feel free to submit further fixups if you need
them.
> > - the ewah code used gcc's __builtin_ctzll, but did not provide a
> > suitable fallback. We now provide a fallback in C.
>
> ... here.
>
> I was messing around with several implementations (including the use of
> msvc compiler intrinsics) with the intention of doing some timing tests
> etc. [I suspected my C fallback function (a different implementation to
> yours) would be slightly faster.]
Yeah, I looked around for several implementations, and ultimately wrote
one that was the most readable to me. The one I found shortest and most
inscrutable was:
return popcount((x & -x) - 1);
the details of which I still haven't worked through in my head. ;)
I do think on most platforms that intrinsics or inline assembler are the
way to go. My main goal was to get something correct that would let it
compile everywhere, and then people can use that as a base for
optimizing. Patches welcome. :)
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 0/21] pack bitmaps
2013-11-14 21:33 ` Jeff King
@ 2013-11-14 23:09 ` Ramsay Jones
2013-11-18 21:16 ` Ramsay Jones
2013-11-16 10:28 ` Thomas Rast
1 sibling, 1 reply; 55+ messages in thread
From: Ramsay Jones @ 2013-11-14 23:09 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
On 14/11/13 21:33, Jeff King wrote:
> On Thu, Nov 14, 2013 at 07:19:38PM +0000, Ramsay Jones wrote:
>
>> Unfortunately, I didn't find time this weekend to finish the msvc build
>> fixes. However, after a quick squint at these patches, I think you have
>> almost done it for me! :-D
>>
>> I must have misunderstood the previous discussion, because my patch was
>> written on the assumption that the ewah directory wouldn't be "git-ified"
>> (e.g. #include git-compat-util.h).
>
> I think it was up for debate at some point, but we did decide to go
> ahead and git-ify. Please feel free to submit further fixups if you need
> them.
Yep, will do; at present it looks like that one-liner.
>
>>> - the ewah code used gcc's __builtin_ctzll, but did not provide a
>>> suitable fallback. We now provide a fallback in C.
>>
>> ... here.
>>
>> I was messing around with several implementations (including the use of
>> msvc compiler intrinsics) with the intention of doing some timing tests
>> etc. [I suspected my C fallback function (a different implementation to
>> yours) would be slightly faster.]
>
> Yeah, I looked around for several implementations, and ultimately wrote
> one that was the most readable to me. The one I found shortest and most
> inscrutable was:
>
> return popcount((x & -x) - 1);
>
> the details of which I still haven't worked through in my head. ;)
Yeah, I stumbled over that one too!
> I do think on most platforms that intrinsics or inline assembler are the
> way to go. My main goal was to get something correct that would let it
> compile everywhere, and then people can use that as a base for
> optimizing. Patches welcome. :)
Indeed, I can happily leave that to another day (or to someone else
more motivated ;-)
ATB,
Ramsay Jones
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 0/21] pack bitmaps
2013-11-14 21:33 ` Jeff King
2013-11-14 23:09 ` Ramsay Jones
@ 2013-11-16 10:28 ` Thomas Rast
1 sibling, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-11-16 10:28 UTC (permalink / raw)
To: Jeff King; +Cc: Ramsay Jones, git, Vicent Martí
Jeff King <peff@peff.net> writes:
>> > - the ewah code used gcc's __builtin_ctzll, but did not provide a
>> > suitable fallback. We now provide a fallback in C.
>>
>> I was messing around with several implementations (including the use of
>> msvc compiler intrinsics) with the intention of doing some timing tests
>> etc. [I suspected my C fallback function (a different implementation to
>> yours) would be slightly faster.]
>
> Yeah, I looked around for several implementations, and ultimately wrote
> one that was the most readable to me. The one I found shortest and most
> inscrutable was:
>
> return popcount((x & -x) - 1);
In two's complement, -x = ~x + 1 [1]. If you have a bunch of 0s at the
end, as in (binary; a=~A etc)
x = abcdef1000
then
~x = ABCDEF0111
~x + 1 = -x = ABCDEF1000
(x&-x) = 0000001000
(x&-x) - 1 = 0000000111
popcount() of that is the number of trailing zeroes you started with.
Please don't ask me to work out what happens in border cases; my head
hurts already.
[1] because x + ~x is all one bits. +1 makes it overflow to 0, so that
x + -x = 0 as it should.
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 0/21] pack bitmaps
2013-11-14 23:09 ` Ramsay Jones
@ 2013-11-18 21:16 ` Ramsay Jones
0 siblings, 0 replies; 55+ messages in thread
From: Ramsay Jones @ 2013-11-18 21:16 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
On 14/11/13 23:09, Ramsay Jones wrote:
> On 14/11/13 21:33, Jeff King wrote:
>> On Thu, Nov 14, 2013 at 07:19:38PM +0000, Ramsay Jones wrote:
>>
>>> Unfortunately, I didn't find time this weekend to finish the msvc build
>>> fixes. However, after a quick squint at these patches, I think you have
>>> almost done it for me! :-D
>>>
>>> I must have misunderstood the previous discussion, because my patch was
>>> written on the assumption that the ewah directory wouldn't be "git-ified"
>>> (e.g. #include git-compat-util.h).
>>
>> I think it was up for debate at some point, but we did decide to go
>> ahead and git-ify. Please feel free to submit further fixups if you need
>> them.
>
> Yep, will do; at present it looks like that one-liner.
Despite saying I would wait for these patches to land in pu, I applied these
patches to the next branch (@ 8721652 "Sync with 1.8.5-rc2") so that I could
try them out.
As expected, everything was fine on Linux and cygwin, and the msvc build needed
a one-liner fix. However, I didn't expect that MinGW would need the same fix! ;-)
I had forgotten that "git-compat-util.h" does not include the <inttypes.h> header
on MinGW (my previous patch included that header directly), so that we have to
add the printf format macros directly to the compat/mingw.h header.
[I assume that an earlier version of the library on MinGW did not have the
<inttypes.h> and <stdint.h> headers. We could now include that header on MinGW
and move the PRIuMAX, PRId64 and PRIx64 macros to compat/msvc.h, but I didn't
think it was worth the churn ... ]
The one-liner is given below ...
ATB,
Ramsay Jones
-- >8 --
Subject: [PATCH] compat/mingw.h: Fix the MinGW and msvc builds
Signed-off-by: Ramsay Jones <ramsay@ramsay1.demon.co.uk>
---
compat/mingw.h | 1 +
1 file changed, 1 insertion(+)
diff --git a/compat/mingw.h b/compat/mingw.h
index 92cd728..8828ede 100644
--- a/compat/mingw.h
+++ b/compat/mingw.h
@@ -345,6 +345,7 @@ static inline char *mingw_find_last_dir_sep(const char *path)
#define PATH_SEP ';'
#define PRIuMAX "I64u"
#define PRId64 "I64d"
+#define PRIx64 "I64x"
void mingw_open_html(const char *path);
#define open_html mingw_open_html
--
1.8.4
^ permalink raw reply related [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-14 12:44 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Jeff King
@ 2013-11-24 21:36 ` Thomas Rast
2013-11-25 15:04 ` [PATCH] Document khash Thomas Rast
2013-11-27 9:08 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Karsten Blees
2013-11-29 21:21 ` Thomas Rast
1 sibling, 2 replies; 55+ messages in thread
From: Thomas Rast @ 2013-11-24 21:36 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> khash.h | 338 ++++++++++++++++++++
[...]
> diff --git a/khash.h b/khash.h
> new file mode 100644
> index 0000000..57ff603
> --- /dev/null
> +++ b/khash.h
> @@ -0,0 +1,338 @@
> +/* The MIT License
> +
> + Copyright (c) 2008, 2009, 2011 by Attractive Chaos <attractor@live.co.uk>
> +
> + Permission is hereby granted, free of charge, to any person obtaining
> + a copy of this software and associated documentation files (the
> + "Software"), to deal in the Software without restriction, including
> + without limitation the rights to use, copy, modify, merge, publish,
> + distribute, sublicense, and/or sell copies of the Software, and to
> + permit persons to whom the Software is furnished to do so, subject to
> + the following conditions:
> +
> + The above copyright notice and this permission notice shall be
> + included in all copies or substantial portions of the Software.
> +
> + THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND,
> + EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
> + MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND
> + NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS
> + BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN
> + ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN
> + CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
> + SOFTWARE.
> +*/
> +
> +#ifndef __AC_KHASH_H
> +#define __AC_KHASH_H
[...]
> +static inline khint_t __kh_oid_hash(const unsigned char *oid)
> +{
> + khint_t hash;
> + memcpy(&hash, oid, sizeof(hash));
> + return hash;
> +}
> +
> +#define __kh_oid_cmp(a, b) (hashcmp(a, b) == 0)
> +
> +KHASH_INIT(sha1, const unsigned char *, void *, 1, __kh_oid_hash, __kh_oid_cmp)
> +typedef kh_sha1_t khash_sha1;
> +
> +KHASH_INIT(sha1_pos, const unsigned char *, int, 1, __kh_oid_hash, __kh_oid_cmp)
> +typedef kh_sha1_pos_t khash_sha1_pos;
> +
> +#endif /* __AC_KHASH_H */
AFAICS, the part after the [...] are additions specific to git. Is that
right?
Can we store them in a separate khash-sha1.h or some such, to make it
clearer what's what? As things stand, one has to look for an identifier
that is built from macros at the far end of a file that looks like it
was imported verbatim from klib(?). I don't know about you, but that
just took me far too long to get right.
I think I'll also lend you a hand writing Documentation/technical/api-khash.txt
(expect it tomorrow) so that we also have documentation in the git
style, where gitters can be expected to find it on their own.
All that could then nicely fit into a commit that actually says where
you conjured khash.h from and such. This information seems
conspicuously absent from this commit's message :-)
Furthermore, would it be a problem to name the second hash sha1_int
instead? I have another use for such a hash, and I can't imagine I'm
the only one. (That's not critical however, I can do the required
editing in that other series.)
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* [PATCH] Document khash
2013-11-24 21:36 ` Thomas Rast
@ 2013-11-25 15:04 ` Thomas Rast
2013-11-28 10:35 ` Jeff King
2013-11-27 9:08 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Karsten Blees
1 sibling, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-11-25 15:04 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
For squashing into a commit that adds khash.h.
Signed-off-by: Thomas Rast <tr@thomasrast.ch>
---
> I think I'll also lend you a hand writing Documentation/technical/api-khash.txt
> (expect it tomorrow) so that we also have documentation in the git
> style, where gitters can be expected to find it on their own.
Here goes.
> Furthermore, would it be a problem to name the second hash sha1_int
> instead? I have another use for such a hash, and I can't imagine I'm
> the only one. (That's not critical however, I can do the required
> editing in that other series.)
Actually, let's not do that. Since everything is 'static inline'
anyway, there's no cost to simply instantiating more hashes as needed.
Documentation/technical/api-khash.txt | 109 ++++++++++++++++++++++++++++++++++
1 file changed, 109 insertions(+)
create mode 100644 Documentation/technical/api-khash.txt
diff --git a/Documentation/technical/api-khash.txt b/Documentation/technical/api-khash.txt
new file mode 100644
index 0000000..e7ca883
--- /dev/null
+++ b/Documentation/technical/api-khash.txt
@@ -0,0 +1,109 @@
+khash
+=====
+
+The khash API is a collection of macros that instantiate hash table
+routines for arbitrary key/value types.
+
+It lives in 'khash.h', which we imported from
+ https://github.com/attractivechaos/klib/blob/master/khash.h
+and then gitified somewhat.
+
+Usage
+-----
+
+------------
+#import "khash.h"
+KHASH_INIT(NAME, key_t, value_t, is_map, key_hash_fn, key_equal_fn)
+------------
+
+The arguments are as follows:
+
+`NAME`::
+ Used to give your hash and its API a unique name. Spelled
+ `NAME` here to set it apart from the rest of the names, but
+ generally should be a lowercase C identifier.
+
+`key_t`::
+ Type of the keys for your hashes. Generally should be
+ `const`.
+
+`value_t`::
+ Type of the values of your hashes.
+
+`is_map`::
+ Whether the hash holds values. Set to 0 to create a hash for
+ use as a set.
+
+`khint_t key_hash_fn(key_t)`::
+ Hash function.
+
+`int key_equal_fn(key_t a, key_t b)`::
+ Comparison function. Return 1 if the two keys are the same, 0
+ otherwise.
+
+These two functions may also be macros.
+
+
+API
+---
+
+The above instantiation defines a single type:
+
+`kh_NAME_t`::
+ A struct that holds your hash. You should only ever use
+ pointers to it.
+
+After the above instantiation, the following functions and
+function-like macros are defined:
+
+`kh_NAME_t *kh_init_NAME(void)`::
+ Allocate and initialize a new hash table.
+
+`void kh_destroy_NAME(kh_NAME_t *)`::
+ Free the hash table.
+
+`void kh_clear_NAME(kh_NAME_t *)`::
+ Clear (but do not free) the hash table.
+
+`khint_t kh_get_NAME(const kh_NAME_t *hash, key_t key)`::
+ Find the given key in the hash table. The returned khint_t
+ should be treated as an opaque iterator token that indexes
+ into the hash. Use `kh_value` to get the value from the
+ token. If the key does not exist, returns kh_end(hash).
+
+`khint_t kh_put_NAME(const kh_NAME_t *hash, key_t key, int *ret)`::
+ Put the given key in the hash table. The returned khint_t
+ should be treated as an opaque iterator token that indexes
+ into the hash. Use `kh_value(hash, token) = value` to assign
+ a value after putting the key.
++
+'ret' tells you whether the key already existed: it is -1 if the
+operation failed; 0 if the key was already present; 1 if the bucket is
+empty; 2 if the bucket has been deleted.
+
+`kh_key(kh_NAME_t *, khint_t)`::
+ The key slot for the given iterator token. This can be used
+ as an lvalue, but you must not change the key!
+
+`kh_value(kh_NAME_t *, khint_t)`::
+`kh_val(kh_NAME_t *, khint_t)`::
+ The value slot for the given iterator token. This can be used
+ as an lvalue.
+
+`int kh_exist(const kh_NAME_t *, key_t)`::
+ Tests if the given bucket contains data.
+
+`khint_t kh_begin(const kh_NAME_t *)`::
+ The smallest iterator token.
+
+`khint_t kh_end(const kh_NAME_t *)`::
+ The beyond-the-hash iterator token; it is one larger than the
+ largest token that points to a hash member.
+
+`khint_t kh_size(const kh_NAME_t *)`::
+ The number of elements in the hash.
+
+`kh_foreach(const kh_NAME_t *, keyvar, valuevar, { ... loop body ... })`::
+ Iterate over every key-value pair in the hash. This is a
+ macro, and keyvar and valuevar must be pre-declared of type
+ key_t and value_t, respectively.
--
1.8.5.rc3.397.g2a3acd5
^ permalink raw reply related [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-24 21:36 ` Thomas Rast
2013-11-25 15:04 ` [PATCH] Document khash Thomas Rast
@ 2013-11-27 9:08 ` Karsten Blees
2013-11-28 10:38 ` Jeff King
1 sibling, 1 reply; 55+ messages in thread
From: Karsten Blees @ 2013-11-27 9:08 UTC (permalink / raw)
To: Thomas Rast, Jeff King; +Cc: git, Vicent Martí
Am 24.11.2013 22:36, schrieb Thomas Rast:
> Jeff King <peff@peff.net> writes:
[...]
>
> I think I'll also lend you a hand writing Documentation/technical/api-khash.txt
> (expect it tomorrow) so that we also have documentation in the git
> style, where gitters can be expected to find it on their own.
>
Khash is OK for sha1 keys, but I don't think it should be advertised as a second general purpose hash table implementation. Its far too easy to shoot yourself in the foot by using 'straightforward' hash- and comparison functions. Khash doesn't store the hash codes of the keys, so you have to take care of that yourself or live with the performance penalties (see [1]).
[1] http://article.gmane.org/gmane.comp.version-control.git/237876
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH] Document khash
2013-11-25 15:04 ` [PATCH] Document khash Thomas Rast
@ 2013-11-28 10:35 ` Jeff King
0 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-11-28 10:35 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Mon, Nov 25, 2013 at 04:04:48PM +0100, Thomas Rast wrote:
> > I think I'll also lend you a hand writing Documentation/technical/api-khash.txt
> > (expect it tomorrow) so that we also have documentation in the git
> > style, where gitters can be expected to find it on their own.
>
> Here goes.
Thanks. Some comments below.
> > Furthermore, would it be a problem to name the second hash sha1_int
> > instead? I have another use for such a hash, and I can't imagine I'm
> > the only one. (That's not critical however, I can do the required
> > editing in that other series.)
>
> Actually, let's not do that. Since everything is 'static inline'
> anyway, there's no cost to simply instantiating more hashes as needed.
Yeah. If we cared about code duplication, we'd have to split it into
declaration and definition macros, and then collect the definitions in
one .c file (khash does support this, but we don't use it). I don't
think we are using enough repeated ones to make it matter, and most of
the functions benefit from inlining anyway.
> +------------
> +#import "khash.h"
> +KHASH_INIT(NAME, key_t, value_t, is_map, key_hash_fn, key_equal_fn)
> +------------
#import?
> +The arguments are as follows:
> [...]
> +`khint_t key_hash_fn(key_t)`::
> + Hash function.
It is true that this is a khint_t, but I do not think knowing that helps
the caller. They need to design a hash function, so knowing the size and
type of the hint helps. Maybe:
Return a hash value for a key. The khint_t is a 32-bit unsigned value.
Git provides __kh_oid_hash, which converts a sha1 into a hash value.
> +`int key_equal_fn(key_t a, key_t b)`::
> + Comparison function. Return 1 if the two keys are the same, 0
> + otherwise.
Here we provide __kh_oid_cmp, and we should mention it. Based on recent
discussions, this should probably be __kh_oid_equal, since it is not a
"cmp" function in the ordering sense.
> +`khint_t kh_get_NAME(const kh_NAME_t *hash, key_t key)`::
> + Find the given key in the hash table. The returned khint_t
> + should be treated as an opaque iterator token that indexes
> + into the hash. Use `kh_value` to get the value from the
> + token. If the key does not exist, returns kh_end(hash).
I do not know of the khash author's intention, but in the bitmap code we
prefer khiter_t to represent an iterator. It is the same as a khint_t,
but I think that is an implementation detail (and the khash functions
should probably be returning a khiter_t).
> [...]
The rest of it looks correct to me.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-27 9:08 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Karsten Blees
@ 2013-11-28 10:38 ` Jeff King
2013-12-03 14:40 ` Karsten Blees
0 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-11-28 10:38 UTC (permalink / raw)
To: Karsten Blees; +Cc: Thomas Rast, git, Vicent Martí
On Wed, Nov 27, 2013 at 10:08:56AM +0100, Karsten Blees wrote:
> Khash is OK for sha1 keys, but I don't think it should be advertised
> as a second general purpose hash table implementation. Its far too
> easy to shoot yourself in the foot by using 'straightforward' hash-
> and comparison functions. Khash doesn't store the hash codes of the
> keys, so you have to take care of that yourself or live with the
> performance penalties (see [1]).
>
> [1] http://article.gmane.org/gmane.comp.version-control.git/237876
Yes. I wonder if we should improve it in that respect. I haven't looked
carefully at the hash code you posted elsewhere, but I feel like many
uses will want a macro implementation to let them store arbitrary types
smaller or larger than a pointer.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-14 12:44 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Jeff King
2013-11-24 21:36 ` Thomas Rast
@ 2013-11-29 21:21 ` Thomas Rast
2013-12-02 16:12 ` Jeff King
1 sibling, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-11-29 21:21 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
TLDR: nitpicks. Thanks for a very nice read.
I do think it's worth fixing the syntax pedantry at the end so that we
can keep supporting arcane compilers, but otherwise, meh.
> +static int open_pack_bitmap_1(struct packed_git *packfile)
This goes somewhat against the naming convention (if you can call it
that) used elsewhere in git. Usually foo_1() is an implementation
detail of foo(), used because it is convenient to wrap the main part in
another function, e.g. so that it can consistently free resources or
some such. But this one operates on one pack file, so in the terms of
the rest of git, it should probably be called open_pack_bitmap_one().
> +static void show_object(struct object *object, const struct name_path *path,
> + const char *last, void *data)
> +{
> + struct bitmap *base = data;
> + int bitmap_pos;
> +
> + bitmap_pos = bitmap_position(object->sha1);
> +
> + if (bitmap_pos < 0) {
> + char *name = path_name(path, last);
> + bitmap_pos = ext_index_add_object(object, name);
> + free(name);
> + }
> +
> + bitmap_set(base, bitmap_pos);
> +}
> +
> +static void show_commit(struct commit *commit, void *data)
> +{
> +}
A bit unfortunate that you inherit the strange show_* naming from
builtin/pack-objects.c, which seems to have stolen some code from
builtin/rev-list.c at some point without worrying about better naming...
> +static void show_objects_for_type(
> + struct bitmap *objects,
> + struct ewah_bitmap *type_filter,
> + enum object_type object_type,
> + show_reachable_fn show_reach)
> +{
[...]
> + while (i < objects->word_alloc && ewah_iterator_next(&filter, &it)) {
> + eword_t word = objects->words[i] & filter;
> +
> + for (offset = 0; offset < BITS_IN_WORD; ++offset) {
> + const unsigned char *sha1;
> + struct revindex_entry *entry;
> + uint32_t hash = 0;
> +
> + if ((word >> offset) == 0)
> + break;
> +
> + offset += ewah_bit_ctz64(word >> offset);
> +
> + if (pos + offset < bitmap_git.reuse_objects)
> + continue;
> +
> + entry = &bitmap_git.reverse_index->revindex[pos + offset];
> + sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
> +
> + show_reach(sha1, object_type, 0, hash, bitmap_git.pack, entry->offset);
> + }
You have a very nice bitmap_each_bit() function in ewah/bitmap.c, why
not use it here?
> +int reuse_partial_packfile_from_bitmap(struct packed_git **packfile,
> + uint32_t *entries,
> + off_t *up_to)
> +{
> + /*
> + * Reuse the packfile content if we need more than
> + * 90% of its objects
> + */
> + static const double REUSE_PERCENT = 0.9;
Curious: is this based on some measurements or just a guess?
> diff --git a/pack-bitmap.h b/pack-bitmap.h
[...]
> +static const char BITMAP_IDX_SIGNATURE[] = {'B', 'I', 'T', 'M'};;
There's a stray ; at the end of the line that is technically not
permitted:
pack-bitmap.h:22:65: warning: ISO C does not allow extra ‘;’ outside of a function [-Wpedantic]
> +enum pack_bitmap_opts {
> + BITMAP_OPT_FULL_DAG = 1,
And I think this trailing comma on the last enum item is also strictly
speaking not allowed, even though it is very nice to have:
pack-bitmap.h:28:27: warning: comma at end of enumerator list [-Wpedantic]
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-29 21:21 ` Thomas Rast
@ 2013-12-02 16:12 ` Jeff King
2013-12-02 20:36 ` Junio C Hamano
0 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-12-02 16:12 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Fri, Nov 29, 2013 at 10:21:04PM +0100, Thomas Rast wrote:
> I do think it's worth fixing the syntax pedantry at the end so that we
> can keep supporting arcane compilers, but otherwise, meh.
Agreed. I've picked up those changes in my tree.
> > +static int open_pack_bitmap_1(struct packed_git *packfile)
>
> This goes somewhat against the naming convention (if you can call it
> that) used elsewhere in git. Usually foo_1() is an implementation
> detail of foo(), used because it is convenient to wrap the main part in
> another function, e.g. so that it can consistently free resources or
> some such. But this one operates on one pack file, so in the terms of
> the rest of git, it should probably be called open_pack_bitmap_one().
Hmm. I see your point, but I think that my (and Vicent's) mental model
was that is _was_ a helper for open_pack_bitmap. It just happens to also
fill the role of open_pack_bitmap_one(), but you would not want the
latter. We only support a single bitmap at a time; by calling the
helper, you would miss out on the assert which would catch the error.
So I don't care much, but I have a slight preference to leave it, as it
signals "you should not be calling this directly" more clearly.
> A bit unfortunate that you inherit the strange show_* naming from
> builtin/pack-objects.c, which seems to have stolen some code from
> builtin/rev-list.c at some point without worrying about better naming...
Yes, I agree they're not very descriptive. Let's leave it for now to
stay consistent with pack-objects, and I'd be happy to see a patch
giving all of them better names come later.
> > + while (i < objects->word_alloc && ewah_iterator_next(&filter, &it)) {
> > + eword_t word = objects->words[i] & filter;
> > +
> > + for (offset = 0; offset < BITS_IN_WORD; ++offset) {
> > + const unsigned char *sha1;
> > + struct revindex_entry *entry;
> > + uint32_t hash = 0;
> > +
> > + if ((word >> offset) == 0)
> > + break;
> > +
> > + offset += ewah_bit_ctz64(word >> offset);
> > +
> > + if (pos + offset < bitmap_git.reuse_objects)
> > + continue;
> > +
> > + entry = &bitmap_git.reverse_index->revindex[pos + offset];
> > + sha1 = nth_packed_object_sha1(bitmap_git.pack, entry->nr);
> > +
> > + show_reach(sha1, object_type, 0, hash, bitmap_git.pack, entry->offset);
> > + }
>
> You have a very nice bitmap_each_bit() function in ewah/bitmap.c, why
> not use it here?
We are bitwise-ANDing against an on-disk ewah bitmap to filter out
objects which do not match the desired type. bitmap_each_bit would make
this more complicated, because we wouldn't be able to move the
ewah_iterator in single-word lockstep. And it would probably be slower
(if you did it naively), because we'd end up checking each bit in the
ewah, rather than AND-ing whole words.
The right, reusable way to do it would probably be to bitmap_and_ewah
the original and the filter together, and then bitmap_each_bit the
result. But you would have to write bitmap_and_ewah first. :)
> > + /*
> > + * Reuse the packfile content if we need more than
> > + * 90% of its objects
> > + */
> > + static const double REUSE_PERCENT = 0.9;
>
> Curious: is this based on some measurements or just a guess?
I think it's mostly a guess.
> > +enum pack_bitmap_opts {
> > + BITMAP_OPT_FULL_DAG = 1,
>
> And I think this trailing comma on the last enum item is also strictly
> speaking not allowed, even though it is very nice to have:
>
> pack-bitmap.h:28:27: warning: comma at end of enumerator list [-Wpedantic]
It's allowed in C99, but was not in C89. I've fixed this site for
consistency with the rest of git. But I wonder how relevant it still is.
The only data points I know of are:
http://article.gmane.org/gmane.comp.version-control.git/145739
and
http://article.gmane.org/gmane.comp.version-control.git/145739
It sounds like an ancient IBM VisualAge is the only reported problem.
And according to IBM, they stopped supporting it 10 years ago (well,
technically we have a few more weeks to hit the 10-year mark):
http://www-01.ibm.com/common/ssi/cgi-bin/ssialias?infotype=an&subtype=ca&supplier=897&appname=IBMLinkRedirect&letternum=ENUS903-227
I do wonder if at some point we should revisit our "do not use any
C99-isms" philosophy. It was very good advice in 2005. I don't know how
good it is over 8 years later (it seems like even ancient systems should
be able to get gcc compiled as a last resort, but maybe there really are
people for whom that is a burden).
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-12-02 16:12 ` Jeff King
@ 2013-12-02 20:36 ` Junio C Hamano
2013-12-02 20:47 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Junio C Hamano @ 2013-12-02 20:36 UTC (permalink / raw)
To: Jeff King; +Cc: Thomas Rast, git, Vicent Martí
Jeff King <peff@peff.net> writes:
> I do wonder if at some point we should revisit our "do not use any
> C99-isms" philosophy. It was very good advice in 2005. I don't know how
> good it is over 8 years later (it seems like even ancient systems should
> be able to get gcc compiled as a last resort, but maybe there really are
> people for whom that is a burden).
Well, we are not kernel where being able to precisely control
generated machine code matters and enforcement of acceptable
compiler versions to achieve that goal is warranted, so I'd prefer
to avoid anything that tells the users "go get a newer gcc".
There are certain things outside C89 that would make our code easier
to read and maintain (e.g. named member initialization of
struct/union, cf. ANSI C99 s6.7.9, just to name one) that I would
love to be able to use in our codebase, but being able to leave an
extra comma at the list of enums is very low on that list. Other
than making a patch unnecessarily verbose when you introduce a new
enum and append it at the end, I do not think it hurts us much (I
may be missing other reasons why you may want to leave an extra
comma at the end, though).
Even that problem can easily be solved by having the "a value of
this enum has to be lower than this" at the end, e.g.
enum object_type {
OBJ_BAD = -1,
OBJ_NONE = 0,
OBJ_COMMIT = 1,
OBJ_TREE = 2,
OBJ_BLOB = 3,
OBJ_TAG = 4,
/* 5 for future expansion */
OBJ_OFS_DELTA = 6,
OBJ_REF_DELTA = 7,
OBJ_ANY,
OBJ_MAX
};
which will never tempt anybody to append at the end, causing the
patch bloat.
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-12-02 20:36 ` Junio C Hamano
@ 2013-12-02 20:47 ` Jeff King
2013-12-02 21:43 ` Junio C Hamano
0 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-12-02 20:47 UTC (permalink / raw)
To: Junio C Hamano; +Cc: Thomas Rast, git, Vicent Martí
On Mon, Dec 02, 2013 at 12:36:34PM -0800, Junio C Hamano wrote:
> Jeff King <peff@peff.net> writes:
>
> > I do wonder if at some point we should revisit our "do not use any
> > C99-isms" philosophy. It was very good advice in 2005. I don't know how
> > good it is over 8 years later (it seems like even ancient systems should
> > be able to get gcc compiled as a last resort, but maybe there really are
> > people for whom that is a burden).
>
> Well, we are not kernel where being able to precisely control
> generated machine code matters and enforcement of acceptable
> compiler versions to achieve that goal is warranted, so I'd prefer
> to avoid anything that tells the users "go get a newer gcc".
Sorry, I was not very clear about what I said. I do not think "go get a
newer gcc" is a good thing to be telling people. But I wonder:
a. if there are actually people on systems that have pre-c99 compilers
in 2013
b. if there are, do they actually _use_ the ancient system compiler,
and not just install gcc as the first step anyway?
In other words, I am questioning whether we would have to tell anybody
"go install gcc" these days. I'm not sure of the best way to answer that
question, though.
> There are certain things outside C89 that would make our code easier
> to read and maintain (e.g. named member initialization of
> struct/union, cf. ANSI C99 s6.7.9, just to name one) that I would
> love to be able to use in our codebase, but being able to leave an
> extra comma at the list of enums is very low on that list.
Yes, I can live without trailing commas. I was musing more on the
general issue (of course, we don't _have_ to take C99 as a whole, and
can pick and choose features that even pre-C99 compilers got right, but
I was wondering mainly when it would be time to say C99 is "old enough"
that everybody supports it).
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-12-02 20:47 ` Jeff King
@ 2013-12-02 21:43 ` Junio C Hamano
0 siblings, 0 replies; 55+ messages in thread
From: Junio C Hamano @ 2013-12-02 21:43 UTC (permalink / raw)
To: Jeff King; +Cc: Thomas Rast, git, Vicent Martí
Jeff King <peff@peff.net> writes:
> Sorry, I was not very clear about what I said. I do not think "go get a
> newer gcc" is a good thing to be telling people. But I wonder:
>
> a. if there are actually people on systems that have pre-c99 compilers
> in 2013
>
> b. if there are, do they actually _use_ the ancient system compiler,
> and not just install gcc as the first step anyway?
>
> In other words, I am questioning whether we would have to tell anybody
> "go install gcc" these days. I'm not sure of the best way to answer that
> question, though.
Thanks. We are at the same wavelength. I do not know what the best
way to answer that question, either. Unleashing such a change to
the wild is one obvious way to do so, and it will give us the most
accurate answer, so it may be the "best way" if we primarily care
about the quality of the answer, but we value the experience of our
direct consumers even more when judging what is best, so...
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-11-28 10:38 ` Jeff King
@ 2013-12-03 14:40 ` Karsten Blees
2013-12-03 18:21 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Karsten Blees @ 2013-12-03 14:40 UTC (permalink / raw)
To: Jeff King; +Cc: Thomas Rast, git, Vicent Martí
Am 28.11.2013 11:38, schrieb Jeff King:
> On Wed, Nov 27, 2013 at 10:08:56AM +0100, Karsten Blees wrote:
>
>> Khash is OK for sha1 keys, but I don't think it should be advertised
>> as a second general purpose hash table implementation. Its far too
>> easy to shoot yourself in the foot by using 'straightforward' hash-
>> and comparison functions. Khash doesn't store the hash codes of the
>> keys, so you have to take care of that yourself or live with the
>> performance penalties (see [1]).
>>
>> [1] http://article.gmane.org/gmane.comp.version-control.git/237876
>
> Yes. I wonder if we should improve it in that respect. I haven't looked
> carefully at the hash code you posted elsewhere, but I feel like many
> uses will want a macro implementation to let them store arbitrary types
> smaller or larger than a pointer.
>
> -Peff
>
IMO, trying to improve khash isn't worth the trouble.
With smaller-than-pointer types, khash _may_ actually save a few bytes compared to hash.[ch] or hashmap.[ch]. E.g. a set of 'int's would be a perfect use case for khash. However, you're using bitmaps for that, and khash's predefined macros for int sets and maps have been removed from the git version.
Using khash with pointers and larger-than-pointer types is just a waste of memory and performance, though. Hash tables are sparsely filled by design, and using large types means lots of large empty buckets. E.g. kh_resize requires almost four times the size or your data, and copies everything at least three times.
Additionally, there are the obvious problems with khash's macro design (hard to read, impossible to debug, and each named instance increases executable size by ~4k).
Below is a patch that converts pack-bitmap.c to hashmap. Its not even longer than the khash version, and the hashmap API forces you to think about no-brainer improvements such as specifying an expected size or skipping duplicates checks where they aren't needed. I could do the same for pack-bitmap-write.c if you like.
Karsten
-----------8<----------
Subject: [PATCH] pack-bitmap.c: convert to new hashmap implementation
Preallocates bitmap_index.bitmaps with the expected size to prevent
rehashing.
Combines the five arrays of ext_index ('objects', 'hashes' and khash's
'keys', 'flags' and 'vals') into a single structure for better locality.
Removes two unnecessary duplicates checks:
- we don't expect a pack index file to contain duplicate sha1's
- ext_index_add_object is only called after bitmap_position returned -1,
so it is impossible that the entry is already there
Signed-off-by: Karsten Blees <blees@dcon.de>
---
pack-bitmap.c | 166 ++++++++++++++++++++++++++++------------------------------
1 file changed, 81 insertions(+), 85 deletions(-)
diff --git a/pack-bitmap.c b/pack-bitmap.c
index 078f7c6..115caed 100644
--- a/pack-bitmap.c
+++ b/pack-bitmap.c
@@ -15,6 +15,7 @@
* commit.
*/
struct stored_bitmap {
+ struct hashmap_entry ent;
unsigned char sha1[20];
struct ewah_bitmap *root;
struct stored_bitmap *xor;
@@ -22,6 +23,16 @@ struct stored_bitmap {
};
/*
+ * An entry in the extended index.
+ */
+struct ext_entry {
+ struct hashmap_entry ent;
+ uint32_t name_hash;
+ struct object *object;
+ unsigned int nr;
+};
+
+/*
* The currently active bitmap index. By design, repositories only have
* a single bitmap index available (the index for the biggest packfile in
* the repository), since bitmap indexes need full closure.
@@ -61,7 +72,7 @@ static struct bitmap_index {
struct ewah_bitmap *tags;
/* Map from SHA1 -> `stored_bitmap` for all the bitmapped comits */
- khash_sha1 *bitmaps;
+ struct hashmap bitmaps;
/* Number of bitmapped commits */
uint32_t entry_count;
@@ -76,12 +87,7 @@ static struct bitmap_index {
* packed in `pack`, these objects are added to this "fake index" and
* are assumed to appear at the end of the packfile for all operations
*/
- struct eindex {
- struct object **objects;
- uint32_t *hashes;
- uint32_t count, alloc;
- khash_sha1_pos *positions;
- } ext_index;
+ struct hashmap ext_index;
/* Bitmap result of the last performed walk */
struct bitmap *result;
@@ -93,6 +99,21 @@ static struct bitmap_index {
} bitmap_git;
+static int stored_bitmap_cmp(const struct stored_bitmap *entry,
+ const struct stored_bitmap *entry_or_key,
+ const unsigned char *sha1)
+{
+ return hashcmp(entry->sha1, sha1 ? sha1 : entry_or_key->sha1);
+}
+
+static int ext_entry_cmp(const struct ext_entry *entry,
+ const struct ext_entry *entry_or_key,
+ const unsigned char *sha1)
+{
+ return hashcmp(entry->object->sha1,
+ sha1 ? sha1 : entry_or_key->object->sha1);
+}
+
static struct ewah_bitmap *lookup_stored_bitmap(struct stored_bitmap *st)
{
struct ewah_bitmap *parent;
@@ -112,6 +133,16 @@ static struct ewah_bitmap *lookup_stored_bitmap(struct stored_bitmap *st)
return composed;
}
+static struct ewah_bitmap *lookup_stored_bitmap_sha1(const unsigned char *sha1)
+{
+ struct stored_bitmap key, *st;
+ hashmap_entry_init(&key, __kh_oid_hash(sha1));
+ st = hashmap_get(&bitmap_git.bitmaps, &key, sha1);
+ if (st)
+ return lookup_stored_bitmap(st);
+ return NULL;
+}
+
/*
* Read a bitmap from the current read position on the mmaped
* index, and increase the read position accordingly
@@ -174,26 +205,15 @@ static struct stored_bitmap *store_bitmap(struct bitmap_index *index,
int flags)
{
struct stored_bitmap *stored;
- khiter_t hash_pos;
- int ret;
stored = xmalloc(sizeof(struct stored_bitmap));
+ hashmap_entry_init(stored, __kh_oid_hash(sha1));
stored->root = root;
stored->xor = xor_with;
stored->flags = flags;
hashcpy(stored->sha1, sha1);
- hash_pos = kh_put_sha1(index->bitmaps, stored->sha1, &ret);
-
- /* a 0 return code means the insertion succeeded with no changes,
- * because the SHA1 already existed on the map. this is bad, there
- * shouldn't be duplicated commits in the index */
- if (ret == 0) {
- error("Duplicate entry in bitmap index: %s", sha1_to_hex(sha1));
- return NULL;
- }
-
- kh_value(index->bitmaps, hash_pos) = stored;
+ hashmap_add(&index->bitmaps, stored);
return stored;
}
@@ -291,8 +311,9 @@ static int load_pack_bitmap(void)
{
assert(bitmap_git.map && !bitmap_git.loaded);
- bitmap_git.bitmaps = kh_init_sha1();
- bitmap_git.ext_index.positions = kh_init_sha1_pos();
+ hashmap_init(&bitmap_git.bitmaps, (hashmap_cmp_fn) stored_bitmap_cmp,
+ bitmap_git.entry_count);
+ hashmap_init(&bitmap_git.ext_index, (hashmap_cmp_fn) ext_entry_cmp, 0);
bitmap_git.reverse_index = revindex_for_pack(bitmap_git.pack);
if (!(bitmap_git.commits = read_bitmap_1(&bitmap_git)) ||
@@ -362,13 +383,12 @@ struct include_data {
static inline int bitmap_position_extended(const unsigned char *sha1)
{
- khash_sha1_pos *positions = bitmap_git.ext_index.positions;
- khiter_t pos = kh_get_sha1_pos(positions, sha1);
+ struct ext_entry key, *e;
+ hashmap_entry_init(&key, __kh_oid_hash(sha1));
+ e = hashmap_get(&bitmap_git.bitmaps, &key, sha1);
- if (pos < kh_end(positions)) {
- int bitmap_pos = kh_value(positions, pos);
- return bitmap_pos + bitmap_git.pack->num_objects;
- }
+ if (e)
+ return e->nr + bitmap_git.pack->num_objects;
return -1;
}
@@ -390,32 +410,13 @@ static int bitmap_position(const unsigned char *sha1)
static int ext_index_add_object(struct object *object, const char *name)
{
- struct eindex *eindex = &bitmap_git.ext_index;
-
- khiter_t hash_pos;
- int hash_ret;
- int bitmap_pos;
-
- hash_pos = kh_put_sha1_pos(eindex->positions, object->sha1, &hash_ret);
- if (hash_ret > 0) {
- if (eindex->count >= eindex->alloc) {
- eindex->alloc = (eindex->alloc + 16) * 3 / 2;
- eindex->objects = xrealloc(eindex->objects,
- eindex->alloc * sizeof(struct object *));
- eindex->hashes = xrealloc(eindex->hashes,
- eindex->alloc * sizeof(uint32_t));
- }
-
- bitmap_pos = eindex->count;
- eindex->objects[eindex->count] = object;
- eindex->hashes[eindex->count] = pack_name_hash(name);
- kh_value(eindex->positions, hash_pos) = bitmap_pos;
- eindex->count++;
- } else {
- bitmap_pos = kh_value(eindex->positions, hash_pos);
- }
-
- return bitmap_pos + bitmap_git.pack->num_objects;
+ struct ext_entry *e = xmalloc(sizeof(struct ext_entry));
+ hashmap_entry_init(e, __kh_oid_hash(object->sha1));
+ e->object = object;
+ e->name_hash = pack_name_hash(name);
+ e->nr = bitmap_git.ext_index.size;
+ hashmap_add(&bitmap_git.ext_index, e);
+ return e->nr + bitmap_git.pack->num_objects;
}
static void show_object(struct object *object, const struct name_path *path,
@@ -443,7 +444,7 @@ static int add_to_include_set(struct include_data *data,
const unsigned char *sha1,
int bitmap_pos)
{
- khiter_t hash_pos;
+ struct ewah_bitmap *bm;
if (data->seen && bitmap_get(data->seen, bitmap_pos))
return 0;
@@ -451,10 +452,9 @@ static int add_to_include_set(struct include_data *data,
if (bitmap_get(data->base, bitmap_pos))
return 0;
- hash_pos = kh_get_sha1(bitmap_git.bitmaps, sha1);
- if (hash_pos < kh_end(bitmap_git.bitmaps)) {
- struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, hash_pos);
- bitmap_or_ewah(data->base, lookup_stored_bitmap(st));
+ bm = lookup_stored_bitmap_sha1(sha1);
+ if (bm) {
+ bitmap_or_ewah(data->base, bm);
return 0;
}
@@ -507,12 +507,9 @@ static struct bitmap *find_objects(struct rev_info *revs,
roots = roots->next;
if (object->type == OBJ_COMMIT) {
- khiter_t pos = kh_get_sha1(bitmap_git.bitmaps, object->sha1);
-
- if (pos < kh_end(bitmap_git.bitmaps)) {
- struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, pos);
- struct ewah_bitmap *or_with = lookup_stored_bitmap(st);
-
+ struct ewah_bitmap *or_with =
+ lookup_stored_bitmap_sha1(object->sha1);
+ if (or_with) {
if (base == NULL)
base = ewah_to_bitmap(or_with);
else
@@ -584,17 +581,15 @@ static struct bitmap *find_objects(struct rev_info *revs,
static void show_extended_objects(struct bitmap *objects,
show_reachable_fn show_reach)
{
- struct eindex *eindex = &bitmap_git.ext_index;
- uint32_t i;
-
- for (i = 0; i < eindex->count; ++i) {
- struct object *obj;
+ struct ext_entry *e;
+ struct hashmap_iter iter;
- if (!bitmap_get(objects, bitmap_git.pack->num_objects + i))
+ for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
+ e = hashmap_iter_next(&iter)) {
+ if (!bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
continue;
- obj = eindex->objects[i];
- show_reach(obj->sha1, obj->type, 0, eindex->hashes[i], NULL, 0);
+ show_reach(e->object->sha1, e->object->type, 0, e->name_hash, NULL, 0);
}
}
@@ -831,7 +826,8 @@ void traverse_bitmap_commit_list(show_reachable_fn show_reachable)
static uint32_t count_object_type(struct bitmap *objects,
enum object_type type)
{
- struct eindex *eindex = &bitmap_git.ext_index;
+ struct ext_entry *e;
+ struct hashmap_iter iter;
uint32_t i = 0, count = 0;
struct ewah_iterator it;
@@ -863,9 +859,10 @@ static uint32_t count_object_type(struct bitmap *objects,
count += ewah_bit_popcount64(word);
}
- for (i = 0; i < eindex->count; ++i) {
- if (eindex->objects[i]->type == type &&
- bitmap_get(objects, bitmap_git.pack->num_objects + i))
+ for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
+ e = hashmap_iter_next(&iter)) {
+ if (e->object->type == type &&
+ bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
count++;
}
@@ -928,7 +925,7 @@ void test_bitmap_walk(struct rev_info *revs)
{
struct object *root;
struct bitmap *result = NULL;
- khiter_t pos;
+ struct ewah_bitmap *bm;
size_t result_popcnt;
struct bitmap_test_data tdata;
@@ -942,12 +939,9 @@ void test_bitmap_walk(struct rev_info *revs)
bitmap_git.version, bitmap_git.entry_count);
root = revs->pending.objects[0].item;
- pos = kh_get_sha1(bitmap_git.bitmaps, root->sha1);
-
- if (pos < kh_end(bitmap_git.bitmaps)) {
- struct stored_bitmap *st = kh_value(bitmap_git.bitmaps, pos);
- struct ewah_bitmap *bm = lookup_stored_bitmap(st);
+ bm = lookup_stored_bitmap_sha1(root->sha1);
+ if (bm) {
fprintf(stderr, "Found bitmap for %s. %d bits / %08x checksum\n",
sha1_to_hex(root->sha1), (int)bm->bit_size, ewah_checksum(bm));
@@ -1020,6 +1014,7 @@ int rebuild_existing_bitmaps(struct packing_data *mapping,
struct bitmap *rebuild;
struct stored_bitmap *stored;
struct progress *progress = NULL;
+ struct hashmap_iter iter;
khiter_t hash_pos;
int hash_ret;
@@ -1049,7 +1044,8 @@ int rebuild_existing_bitmaps(struct packing_data *mapping,
if (show_progress)
progress = start_progress("Reusing bitmaps", 0);
- kh_foreach_value(bitmap_git.bitmaps, stored, {
+ for (stored = hashmap_iter_first(&bitmap_git.bitmaps, &iter); stored;
+ stored = hashmap_iter_next(&iter)) {
if (stored->flags & BITMAP_FLAG_REUSE) {
if (!rebuild_bitmap(reposition,
lookup_stored_bitmap(stored),
@@ -1063,7 +1059,7 @@ int rebuild_existing_bitmaps(struct packing_data *mapping,
bitmap_reset(rebuild);
display_progress(progress, ++i);
}
- });
+ }
stop_progress(&progress);
--
^ permalink raw reply related [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-12-03 14:40 ` Karsten Blees
@ 2013-12-03 18:21 ` Jeff King
2013-12-07 20:52 ` Karsten Blees
0 siblings, 1 reply; 55+ messages in thread
From: Jeff King @ 2013-12-03 18:21 UTC (permalink / raw)
To: Karsten Blees; +Cc: Thomas Rast, git, Vicent Martí
On Tue, Dec 03, 2013 at 03:40:41PM +0100, Karsten Blees wrote:
> IMO, trying to improve khash isn't worth the trouble.
>
> With smaller-than-pointer types, khash _may_ actually save a few bytes
> compared to hash.[ch] or hashmap.[ch]. E.g. a set of 'int's would be a
> perfect use case for khash. However, you're using bitmaps for that,
> and khash's predefined macros for int sets and maps have been removed
> from the git version.
True, we are not using it for smaller-than-pointer sizes here. So it may
not be worth thinking about (I was considering more for the general
case). In most instances, though, we can shove the int bits into a
pointer with the right casting. So it probably isn't worth worrying
about (you may waste a few bytes, but probably not more than one word
per entry).
> Using khash with pointers and larger-than-pointer types is just a
> waste of memory and performance, though. Hash tables are sparsely
> filled by design, and using large types means lots of large empty
> buckets. E.g. kh_resize requires almost four times the size or your
> data, and copies everything at least three times.
I think the analysis is more complicated than that, and depends on what
you are storing. If you are storing something that is 1.5 times the size
of a pointer, it is more space efficient to just stick it in the hash
table than it is to have a separate pointer in the hash table (you pay
the load factor penalty only on the single pointer, but you've almost
doubled your total storage).
But again, that's a more general argument. We're not storing anything of
that size here, and in fact I think we are just storing pointers
everywhere.
> Additionally, there are the obvious problems with khash's macro design
> (hard to read, impossible to debug, and each named instance increases
> executable size by ~4k).
Yes, those are all downsides to macros. Type safety is one of the
upsides, though.
Besides macros, I think the major difference between the two
implementations is open-addressing versus chaining. Especially for sets,
we've had good experiences with open-addressing by keeping the load
factor low (e.g., the one in object.c). As you note, the resizing
operation pays some penalty, but in most of our workloads it's largely
irrelevant compared to lookup times.
Chaining typically adds an extra layer of pointer-following to the
lookup, and I'd be worried about that loss of locality. It's true that
when you are storing a pointer you are already hurting locality to
follow the pointer to the key, but I don't know whether that means an
_extra_ layer doesn't still hurt more (and I really mean I don't know --
I'd be interested to see measurements).
In your implementation, it looks like you break even there because you
store the hash directly in the entry, and do a single-word compare (so
you avoid having to follow a pointer to the key in the common case
during lookup). But that also means you're paying extra to store the
hash. That probably makes sense for things like strings, where it takes
some effort to calculate the hash. But not necessarily for sha1s, where
looking at the hash is the same thing as looking at the key bytes (so
you are storing extra bytes, and when you do have a hit on the stored
hash, which is just the first bytes of the sha1, you end up comparing
them again as part of the hashcmp. The tradeoff is that in the
non-matching cases, you avoid an extra level of indirection).
All of these are things I could see helping or hurting, depending on the
case. I'd really love to see more numbers. I tried timing your patch
below, but there was no interesting change. Mostly because the code path
you changed is not heavily exercised (the "ext_index" is for objects
that are not in the pack, and frequent packing keeps that number low).
I'd be really interested to see if you could make the hash in object.c
faster. That one is a prominent lookup bottle-neck (e.g., for "git
rev-list --objects --all"); if you can make that faster, I would be very
convinced that your implementation is fast (note that I am not implying
the converse; if you cannot make it faster, that does not necessarily
mean your implementation sucks, but perhaps only that the existing one
has lots of type-specific optimizations which add up).
Khash also has a lot of bloat (e.g., flags) that the one in object.c
does not have. If you do not care about deletion and are storing
something with a sentinel value (e.g., NULL for pointers), you can trim
quite a bit of fat.
> Below is a patch that converts pack-bitmap.c to hashmap. Its not even
> longer than the khash version, and the hashmap API forces you to think
> about no-brainer improvements such as specifying an expected size or
> skipping duplicates checks where they aren't needed. I could do the
> same for pack-bitmap-write.c if you like.
If it's not too much trouble, I'd be curious to measure the performance
impact on pack-bitmap-write.
> Removes two unnecessary duplicates checks:
> - we don't expect a pack index file to contain duplicate sha1's
We don't expect them to, but it has happened (and caused bugs not too
long ago). What happens after your patch when there are duplicates?
> +static struct ewah_bitmap *lookup_stored_bitmap_sha1(const unsigned char *sha1)
> +{
> + struct stored_bitmap key, *st;
> + hashmap_entry_init(&key, __kh_oid_hash(sha1));
> + st = hashmap_get(&bitmap_git.bitmaps, &key, sha1);
> + if (st)
> + return lookup_stored_bitmap(st);
> + return NULL;
> +}
This interface looks odd to me. You create a fake stored_bitmap for the
hashmap_entry part of it, and then fill in the "hash" field by hashing
the sha1. And then pass the same sha1 in. I guess you are trying to
avoid the hash table knowing about the hash function at all, since it
just stores the hash for each entry already.
I guess that makes sense, but would not work for a system where you
wanted to get rid of the extra hash storage (but as I implied above, I
am not sure if it is helping or hurting for the sha1 case).
> - hash_pos = kh_put_sha1_pos(eindex->positions, object->sha1, &hash_ret);
> - if (hash_ret > 0) {
> - if (eindex->count >= eindex->alloc) {
> - eindex->alloc = (eindex->alloc + 16) * 3 / 2;
> - eindex->objects = xrealloc(eindex->objects,
> - eindex->alloc * sizeof(struct object *));
> - eindex->hashes = xrealloc(eindex->hashes,
> - eindex->alloc * sizeof(uint32_t));
> - }
> -
> - bitmap_pos = eindex->count;
> - eindex->objects[eindex->count] = object;
> - eindex->hashes[eindex->count] = pack_name_hash(name);
> - kh_value(eindex->positions, hash_pos) = bitmap_pos;
> - eindex->count++;
> - } else {
> - bitmap_pos = kh_value(eindex->positions, hash_pos);
> - }
> -
> - return bitmap_pos + bitmap_git.pack->num_objects;
> + struct ext_entry *e = xmalloc(sizeof(struct ext_entry));
> + hashmap_entry_init(e, __kh_oid_hash(object->sha1));
> + e->object = object;
> + e->name_hash = pack_name_hash(name);
> + e->nr = bitmap_git.ext_index.size;
> + hashmap_add(&bitmap_git.ext_index, e);
> + return e->nr + bitmap_git.pack->num_objects;
One of the side effects of the current system is that the array
effectively works as a custom allocator, and we do not pay per-entry
malloc overhead. As I mentioned above, this is not that
heavily-exercised a code path, so it may not matter (and you can convert
it by explicitly using a custom allocator, at which point you could drop
the e->nr field entirely, I'd think).
> @@ -584,17 +581,15 @@ static struct bitmap *find_objects(struct rev_info *revs,
> static void show_extended_objects(struct bitmap *objects,
> show_reachable_fn show_reach)
> {
> - struct eindex *eindex = &bitmap_git.ext_index;
> - uint32_t i;
> -
> - for (i = 0; i < eindex->count; ++i) {
> - struct object *obj;
> + struct ext_entry *e;
> + struct hashmap_iter iter;
>
> - if (!bitmap_get(objects, bitmap_git.pack->num_objects + i))
> + for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
> + e = hashmap_iter_next(&iter)) {
> + if (!bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
> continue;
>
> - obj = eindex->objects[i];
> - show_reach(obj->sha1, obj->type, 0, eindex->hashes[i], NULL, 0);
> + show_reach(e->object->sha1, e->object->type, 0, e->name_hash, NULL, 0);
Before we were iterating in eindex order. Now we are iterating in
hashmap order. That effects our traversal order, and ultimately our
output for something like "rev-list". We'd want to sort on e->nr (or
again, a custom allocator would make this go away because we would
iterate over the array).
> - for (i = 0; i < eindex->count; ++i) {
> - if (eindex->objects[i]->type == type &&
> - bitmap_get(objects, bitmap_git.pack->num_objects + i))
> + for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
> + e = hashmap_iter_next(&iter)) {
> + if (e->object->type == type &&
> + bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
> count++;
Ditto here, but I do not think the order matters in this instance.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 11/21] pack-objects: use bitmaps when packing objects
2013-11-14 12:45 ` [PATCH v3 11/21] pack-objects: use bitmaps when packing objects Jeff King
@ 2013-12-07 15:47 ` Thomas Rast
2013-12-21 13:15 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 15:47 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
This week's nits...
I found this harder to read than the previous patch, but I think it's
mostly because the existing code is already a bit tangled. I think the
second item below is worth fixing, though.
Jeff King <peff@peff.net> writes:
> +static off_t write_reused_pack(struct sha1file *f)
> +{
> + uint8_t buffer[8192];
We usually just call this 'unsigned char'. I can see why this would be
more portable, but git would already fall apart badly on an architecture
where char is not 8 bits.
> + off_t to_write;
> + int fd;
> +
> + if (!is_pack_valid(reuse_packfile))
> + return 0;
> +
> + fd = git_open_noatime(reuse_packfile->pack_name);
> + if (fd < 0)
> + return 0;
> +
> + if (lseek(fd, sizeof(struct pack_header), SEEK_SET) == -1) {
> + close(fd);
> + return 0;
> + }
You do an error return if any of the syscalls in this routine fails, but
there is only one caller and it immediately dies:
} + packfile_size = write_reused_pack(f);
} + if (!packfile_size)
} + die_errno("failed to re-use existing pack");
So if you just died here, when the error happens, you could take the
chance to tell the user _which_ syscall failed.
> +
> + if (reuse_packfile_offset < 0)
> + reuse_packfile_offset = reuse_packfile->pack_size - 20;
> +
> + to_write = reuse_packfile_offset - sizeof(struct pack_header);
> +
> + while (to_write) {
> + int read_pack = xread(fd, buffer, sizeof(buffer));
> +
> + if (read_pack <= 0) {
> + close(fd);
> + return 0;
Similar to the above, but this one may also clobber the 'errno' during
close(), which can lead to misleading messages.
> + }
> +
> + if (read_pack > to_write)
> + read_pack = to_write;
> +
> + sha1write(f, buffer, read_pack);
Not your fault, but sha1write() is an odd function -- it purportedly is
int sha1write(struct sha1file *f, const void *buf, unsigned int count);
but it can only return 0. This goes back all the way to c38138c
(git-pack-objects: write the pack files with a SHA1 csum, 2005-06-26).
> + to_write -= read_pack;
> + }
> +
> + close(fd);
> + written += reuse_packfile_objects;
> + return reuse_packfile_offset - sizeof(struct pack_header);
> +}
[...]
> -static int add_object_entry(const unsigned char *sha1, enum object_type type,
> - const char *name, int exclude)
> +static int add_object_entry_1(const unsigned char *sha1, enum object_type type,
> + int flags, uint32_t name_hash,
> + struct packed_git *found_pack, off_t found_offset)
> {
[...]
> - for (p = packed_git; p; p = p->next) {
> - off_t offset = find_pack_entry_one(sha1, p);
> - if (offset) {
> - if (!found_pack) {
> - if (!is_pack_valid(p)) {
> - warning("packfile %s cannot be accessed", p->pack_name);
> - continue;
> + if (!found_pack) {
> + for (p = packed_git; p; p = p->next) {
> + off_t offset = find_pack_entry_one(sha1, p);
> + if (offset) {
> + if (!found_pack) {
> + if (!is_pack_valid(p)) {
> + warning("packfile %s cannot be accessed", p->pack_name);
> + continue;
> + }
> + found_offset = offset;
> + found_pack = p;
> }
> - found_offset = offset;
> - found_pack = p;
> + if (exclude)
> + break;
> + if (incremental)
> + return 0;
> + if (local && !p->pack_local)
> + return 0;
> + if (ignore_packed_keep && p->pack_local && p->pack_keep)
> + return 0;
> }
> - if (exclude)
> - break;
> - if (incremental)
> - return 0;
> - if (local && !p->pack_local)
> - return 0;
> - if (ignore_packed_keep && p->pack_local && p->pack_keep)
> - return 0;
> }
> }
This function makes my head spin, and you're indenting it yet another
level.
If it's not too much work, can you split it into the three parts that it
really is? IIUC it boils down to
do we have this already?
possibly apply 'exclude', then return
are we coming from a call path that doesn't tell us which pack to take
it from?
find _all_ instances in packs
check if any of them are local .keep packs
if so, return
construct a packlist entry to taste
> entry = packlist_alloc(&to_pack, sha1, index_pos);
> - entry->hash = hash;
> + entry->hash = name_hash;
> if (type)
> entry->type = type;
> if (exclude)
> entry->preferred_base = 1;
> else
> nr_result++;
> +
> + if (flags & OBJECT_ENTRY_NO_TRY_DELTA)
> + entry->no_try_delta = 1;
> +
> if (found_pack) {
> entry->in_pack = found_pack;
> entry->in_pack_offset = found_offset;
> @@ -859,10 +932,21 @@ static int add_object_entry(const unsigned char *sha1, enum object_type type,
>
> display_progress(progress_state, to_pack.nr_objects);
>
> + return 1;
> +}
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists
2013-11-14 12:45 ` [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists Jeff King
@ 2013-12-07 16:05 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:05 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> Documentation/git-rev-list.txt | 1 +
> Documentation/rev-list-options.txt | 8 ++++++++
> builtin/rev-list.c | 39 ++++++++++++++++++++++++++++++++++++++
> 3 files changed, 48 insertions(+)
Nice and short. Cheating though, as a lot of the support code is from
[10/21] ;-)
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 13/21] pack-objects: implement bitmap writing
2013-11-14 12:45 ` [PATCH v3 13/21] pack-objects: implement bitmap writing Jeff King
@ 2013-12-07 16:32 ` Thomas Rast
2013-12-21 13:17 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:32 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
You could fix this:
> +pack.writebitmaps::
> + When true, git will write a bitmap index when packing all
> + objects to disk (e.g., as when `git repack -a` is run). This
^^
Doesn't sound right in my ears. Remove the "as"?
> + index can speed up the "counting objects" phase of subsequent
> + packs created for clones and fetches, at the cost of some disk
> + space and extra time spent on the initial repack. Defaults to
> + false.
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts)
2013-11-14 12:45 ` [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts) Jeff King
@ 2013-12-07 16:34 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:34 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> We have a static array of extensions, but hardcode the size
> of the array in our loops. Let's pull out this magic number,
> which will make it easier to change.
>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
(Ok, this one was easy.)
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 15/21] repack: turn exts array into array-of-struct
2013-11-14 12:46 ` [PATCH v3 15/21] repack: turn exts array into array-of-struct Jeff King
@ 2013-12-07 16:34 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:34 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> This is slightly more verbose, but will let us annotate the
> extensions with further options in future commits.
>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 16/21] repack: handle optional files created by pack-objects
2013-11-14 12:46 ` [PATCH v3 16/21] repack: handle optional files created by pack-objects Jeff King
@ 2013-12-07 16:35 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:35 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> We ask pack-objects to pack to a set of temporary files, and
> then rename them into place. Some files that pack-objects
> creates may be optional (like a .bitmap file), in which case
> we would not want to call rename(). We already call stat()
> and make the chmod optional if the file cannot be accessed.
> We could simply skip the rename step in this case, but that
> would be a minor regression in noticing problems with
> non-optional files (like the .pack and .idx files).
>
> Instead, we can now annotate extensions as optional, and
> skip them if they don't exist (and otherwise rely on
> rename() to barf).
>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 17/21] repack: consider bitmaps when performing repacks
2013-11-14 12:46 ` [PATCH v3 17/21] repack: consider bitmaps when performing repacks Jeff King
@ 2013-12-07 16:37 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:37 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> From: Vicent Marti <tanoku@gmail.com>
>
> Since `pack-objects` will write a `.bitmap` file next to the `.pack` and
> `.idx` files, this commit teaches `git-repack` to consider the new
> bitmap indexes (if they exist) when performing repack operations.
>
> This implies moving old bitmap indexes out of the way if we are
> repacking a repository that already has them, and moving the newly
> generated bitmap indexes into the `objects/pack` directory, next to
> their corresponding packfiles.
>
> Since `git repack` is now capable of handling these `.bitmap` files,
> a normal `git gc` run on a repository that has `pack.writebitmaps` set
> to true in its config file will generate bitmap indexes as part of the
> garbage collection process.
>
> Alternatively, `git repack` can be called with the `-b` switch to
> explicitly generate bitmap indexes if you are experimenting
> and don't want them on all the time.
>
> Signed-off-by: Vicent Marti <tanoku@gmail.com>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking
2013-11-14 12:46 ` [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking Jeff King
@ 2013-12-07 16:38 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:38 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> From: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
>
> Count-objects will report any "garbage" files in the packs
> directory, including files whose extensions it does not
> know (case 1), and files whose matching ".pack" file is
> missing (case 2). Without having learned about ".bitmap"
> files, the current code reports all such files as garbage
> (case 1), even if their pack exists. Instead, they should be
> treated as case 2.
>
> Signed-off-by: Nguyễn Thái Ngọc Duy <pclouds@gmail.com>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 19/21] t: add basic bitmap functionality tests
2013-11-14 12:46 ` [PATCH v3 19/21] t: add basic bitmap functionality tests Jeff King
@ 2013-12-07 16:43 ` Thomas Rast
2013-12-21 13:22 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:43 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> Now that we can read and write bitmaps, we can exercise them
> with some basic functionality tests. These tests aren't
> particularly useful for seeing the benefit, as the test
> repo is too small for it to make a difference. However, we
> can at least check that using bitmaps does not break anything.
>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
One nit:
> +test_expect_success JGIT 'jgit can read our bitmaps' '
> + git clone . compat-us.git &&
> + (
> + cd compat-us.git &&
The name suggests a bare repo, but it is a full clone. Not that it
matters.
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 20/21] t/perf: add tests for pack bitmaps
2013-11-14 12:48 ` [PATCH v3 20/21] t/perf: add tests for pack bitmaps Jeff King
@ 2013-12-07 16:51 ` Thomas Rast
2013-12-21 13:40 ` Jeff King
0 siblings, 1 reply; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:51 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> +test_perf 'simulated fetch' '
> + have=$(git rev-list HEAD --until=1.week.ago -1) &&
This will give you HEAD if your GIT_PERF_LARGE_REPO hasn't seen any
activity lately. I'd prefer something that always takes a fixed commit,
e.g. HEAD~1000, keeping the perf test reproducible over time (not over
changing GIT_PERF_LARGE_REPO, of course).
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache
2013-11-14 12:48 ` [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache Jeff King
@ 2013-12-07 16:59 ` Thomas Rast
0 siblings, 0 replies; 55+ messages in thread
From: Thomas Rast @ 2013-12-07 16:59 UTC (permalink / raw)
To: Jeff King; +Cc: git, Vicent Martí
Jeff King <peff@peff.net> writes:
> Test origin/master HEAD^ HEAD
> -------------------------------------------------------------------------------------------------
> 5310.2: repack to disk 36.81(37.82+1.43) 47.70(48.74+1.41) +29.6% 47.75(48.70+1.51) +29.7%
> 5310.3: simulated clone 30.78(29.70+2.14) 1.08(0.97+0.10) -96.5% 1.07(0.94+0.12) -96.5%
> 5310.4: simulated fetch 3.16(6.10+0.08) 3.54(10.65+0.06) +12.0% 1.70(3.07+0.06) -46.2%
> 5310.6: partial bitmap 36.76(43.19+1.81) 6.71(11.25+0.76) -81.7% 4.08(6.26+0.46) -88.9%
>
> You can see that the time spent on an incremental fetch goes
> down, as our delta heuristics are able to do their work.
> And we save time on the partial bitmap clone for the same
> reason.
The time now goes down across the board compared to master. Good job!
> Signed-off-by: Vicent Marti <tanoku@gmail.com>
> Signed-off-by: Jeff King <peff@peff.net>
Reviewed-by: Thomas Rast <tr@thomasrast.ch>
--
Thomas Rast
tr@thomasrast.ch
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes
2013-12-03 18:21 ` Jeff King
@ 2013-12-07 20:52 ` Karsten Blees
0 siblings, 0 replies; 55+ messages in thread
From: Karsten Blees @ 2013-12-07 20:52 UTC (permalink / raw)
To: Jeff King; +Cc: Thomas Rast, git, Vicent Martí
Am 03.12.2013 19:21, schrieb Jeff King:
> On Tue, Dec 03, 2013 at 03:40:41PM +0100, Karsten Blees wrote:
>
>> IMO, trying to improve khash isn't worth the trouble.
>>
>> With smaller-than-pointer types, khash _may_ actually save a few bytes
>> compared to hash.[ch] or hashmap.[ch]. E.g. a set of 'int's would be a
>> perfect use case for khash. However, you're using bitmaps for that,
>> and khash's predefined macros for int sets and maps have been removed
>> from the git version.
>
> True, we are not using it for smaller-than-pointer sizes here. So it may
> not be worth thinking about (I was considering more for the general
> case). In most instances, though, we can shove the int bits into a
> pointer with the right casting. So it probably isn't worth worrying
> about (you may waste a few bytes, but probably not more than one word
> per entry).
>
>> Using khash with pointers and larger-than-pointer types is just a
>> waste of memory and performance, though. Hash tables are sparsely
>> filled by design, and using large types means lots of large empty
>> buckets. E.g. kh_resize requires almost four times the size or your
>> data, and copies everything at least three times.
>
> I think the analysis is more complicated than that, and depends on what
> you are storing. If you are storing something that is 1.5 times the size
> of a pointer, it is more space efficient to just stick it in the hash
> table than it is to have a separate pointer in the hash table (you pay
> the load factor penalty only on the single pointer, but you've almost
> doubled your total storage).
>
> But again, that's a more general argument. We're not storing anything of
> that size here, and in fact I think we are just storing pointers
> everywhere.
>
Yes. With pointers, its a very close call regarding space. Let's take bitmap_index.bitmaps / struct stored_bitmap as an example. I'm assuming 8 byte pointers and an average load factor of 0.5.
struct stored_bitmap is already kind of an entry structure (key + value) allocated on the heap. This is a very common szenario, all hash.[ch] -> hashmap.[ch] conversions were that way.
The average khash memory requirement per entry is (on top of the data on the heap):
(key + value + flags) / load-factor = (8 + 8 + .25) / .5 = 32.5 bytes
For the hashmap conversion, I simply injected struct hashmap_entry into the existing structure, adding 12 bytes per entry. So the total hashmap memory per entry is:
hashmap_entry + (entry-pointer / load-factor) = 12 + (8 / .5) = 28 bytes
>> Additionally, there are the obvious problems with khash's macro design
>> (hard to read, impossible to debug, and each named instance increases
>> executable size by ~4k).
>
> Yes, those are all downsides to macros. Type safety is one of the
> upsides, though.
>
Well, khash_sha1 maps to void *, so I guess its not that important :-) But if you care about type safety, a common core implementation with a few small wrapper macros would be hugely preferable, don't you think?
> Besides macros, I think the major difference between the two
> implementations is open-addressing versus chaining. Especially for sets,
> we've had good experiences with open-addressing by keeping the load
> factor low (e.g., the one in object.c). As you note, the resizing
> operation pays some penalty, but in most of our workloads it's largely
> irrelevant compared to lookup times.
>
Resizing is reasonably fast and straightforward for both open addressing and chaining. My 'copy three times' remark above was specific to the rehash-in-place stunt in kh_resize (if you haven't wondered what these 64 multi-statement-lines do, you probably don't wanna know... ;-)
But its still best if you know the target size and can prevent resizing altogether.
> Chaining typically adds an extra layer of pointer-following to the
> lookup, and I'd be worried about that loss of locality. It's true that
> when you are storing a pointer you are already hurting locality to
> follow the pointer to the key, but I don't know whether that means an
> _extra_ layer doesn't still hurt more (and I really mean I don't know --
> I'd be interested to see measurements).
>
The locality advantage of open addressing _only_ applies when storing the data directly in the table, but AFAIK we're not doing that anywhere in git. Compared to open addressing with pointers, chaining in fact has _better_ locality.
Iterating a linked list (i.e. chaining) accesses exactly one (1.0) memory location per element.
Iterating an array of pointers (i.e. open addressing with linear probing) accesses ~1.125 memory locations per element (assuming 8 byte pointers and 64 byte cache lines, array[8] will be in a different cache line than array[0], thus the +.125).
Of course, any implementation can add more pointer-following to that minimum. E.g. khash uses separate flags and keys arrays and triangular numbers as probe sequence, so its up to 3 memory locations per element.
Hashmap, on the other hand, allows you to inject it's per-entry data into the user data structure to prevent locality penalties (well, the user data structure could add more pointer-following, but that's not hashmap's fault).
> In your implementation, it looks like you break even there because you
> store the hash directly in the entry, and do a single-word compare (so
> you avoid having to follow a pointer to the key in the common case
> during lookup).
More importantly, it saves calling a function pointer with three parameters unless its a _very_ probable match.
> But that also means you're paying extra to store the
> hash. That probably makes sense for things like strings, where it takes
> some effort to calculate the hash. But not necessarily for sha1s, where
> looking at the hash is the same thing as looking at the key bytes (so
> you are storing extra bytes,
True. Its 4 bytes overhead for sha1s and probably ints (but even for ints you may want to wiggle the bits for better distribution). However, its an utterly necessary optimization for e.g. strings with common prefixes. Think of a Java project where all paths start with com/mycompany/myou/productname/component/... i.e. strcmp has to compare ~40 bytes before finding a difference.
> and when you do have a hit on the stored
> hash, which is just the first bytes of the sha1, you end up comparing
> them again as part of the hashcmp.
Not necessarily - I could skip the part that I've used as hash code...
> The tradeoff is that in the
> non-matching cases, you avoid an extra level of indirection).
>
> All of these are things I could see helping or hurting, depending on the
> case. I'd really love to see more numbers. I tried timing your patch
> below, but there was no interesting change. Mostly because the code path
> you changed is not heavily exercised (the "ext_index" is for objects
> that are not in the pack, and frequent packing keeps that number low).
>
> I'd be really interested to see if you could make the hash in object.c
> faster. That one is a prominent lookup bottle-neck (e.g., for "git
> rev-list --objects --all"); if you can make that faster, I would be very
> convinced that your implementation is fast (note that I am not implying
> the converse; if you cannot make it faster, that does not necessarily
> mean your implementation sucks, but perhaps only that the existing one
> has lots of type-specific optimizations which add up).
>
Ok, let's see...
Hash tables only compare for equality, not for sorting, so the first thing that comes to mind is to compare 4 bytes at a time.
The hashmap version cuts quite a bit of code, but uses slightly more memory (~6 bytes per entry). Still todo: get_indexed_object now only returns the list heads, so callers need to follow the chain...
Then I wrote a custom chaining version (i.e. no cached hash code, inlined hashcmp() etc.). The increased load-factor of 0.8 should fully compensate the additional next pointer in struct object. Ditto for get_indexed_object.
Finally, just for fun, the khash version...
Your move-to-front optimization seems highly specialized to lookup_object access patterns, and also breaks the assumption that concurrent lookup is thread safe. So I didn't bother to pimp hashmap and khash internals, and instead did the other measurements with and without move-to-front enabled. The hashmap move-to-front is via the existing API (remove + add), i.e. its true LRU instead of swap (same for custom-chaining).
Numbers are in seconds, best of 10 runs of 'git rev-list --all --objects >/dev/null' on the linux kernel repo.
fast | move || next | hash | custom | khash |
hash | to || | map | | |
cmp | front || (open) |(chain) |(chain) | (open) |
=======+=======++========+========+========+========+
no | no || 45.593 | 42.924 | 41.033 | 62.974 |
no | yes || 41.017 | 43.095 | 40.478 | |
yes | no || 43.172 | 40.965 | 39.564 | |
yes | yes || 39.332 | 41.008 | 38.869 | |
The fastest versions of each variant can be found here:
https://github.com/kblees/git/commits/kb/optimize-lookup-object-next
https://github.com/kblees/git/commits/kb/optimize-lookup-object-hashmap
https://github.com/kblees/git/commits/kb/optimize-lookup-object-custom-chaining
https://github.com/kblees/git/commits/kb/optimize-lookup-object-khash
> Khash also has a lot of bloat (e.g., flags) that the one in object.c
> does not have. If you do not care about deletion and are storing
> something with a sentinel value (e.g., NULL for pointers), you can trim
> quite a bit of fat.
>
Actually, the lack of deletion in hash.[ch] was the reason I started this...one advantage of chaining is that delete can be implemented efficiently (O(1)) without affecting lookup / insert performance.
>> Below is a patch that converts pack-bitmap.c to hashmap. Its not even
>> longer than the khash version, and the hashmap API forces you to think
>> about no-brainer improvements such as specifying an expected size or
>> skipping duplicates checks where they aren't needed. I could do the
>> same for pack-bitmap-write.c if you like.
>
> If it's not too much trouble, I'd be curious to measure the performance
> impact on pack-bitmap-write.
>
I'll see what I can do
>> Removes two unnecessary duplicates checks:
>> - we don't expect a pack index file to contain duplicate sha1's
>
> We don't expect them to, but it has happened (and caused bugs not too
> long ago). What happens after your patch when there are duplicates?
>
Both entries are added, and hashmap_get returns one of them at random. Re-adding duplicates checks is simple, though:
- hashmap_add(...);
+ if (hashmap_put(...))
+ die("foo has duplicate o's");
>> +static struct ewah_bitmap *lookup_stored_bitmap_sha1(const unsigned char *sha1)
>> +{
>> + struct stored_bitmap key, *st;
>> + hashmap_entry_init(&key, __kh_oid_hash(sha1));
>> + st = hashmap_get(&bitmap_git.bitmaps, &key, sha1);
>> + if (st)
>> + return lookup_stored_bitmap(st);
>> + return NULL;
>> +}
>
> This interface looks odd to me. You create a fake stored_bitmap for the
> hashmap_entry part of it, and then fill in the "hash" field by hashing
> the sha1. And then pass the same sha1 in. I guess you are trying to
> avoid the hash table knowing about the hash function at all, since it
> just stores the hash for each entry already.
>
The general case would be:
struct stored_bitmap key;
/* fill in hash code */
hashmap_entry_init(&key, __kh_oid_hash(sha1));
/* fill in key data */
hashcpy(key.sha1, sha1);
key.another_key_field = foobar;
...
return hashmap_get(&bitmap_git.bitmaps, &key, NULL);
In this case, the sha1 is the only key data, and its larger that a word, so passing it as keydata parameter and checking for this in the cmp-function is slightly more efficient.
> I guess that makes sense, but would not work for a system where you
> wanted to get rid of the extra hash storage (but as I implied above, I
> am not sure if it is helping or hurting for the sha1 case).
>
>> - hash_pos = kh_put_sha1_pos(eindex->positions, object->sha1, &hash_ret);
>> - if (hash_ret > 0) {
>> - if (eindex->count >= eindex->alloc) {
>> - eindex->alloc = (eindex->alloc + 16) * 3 / 2;
>> - eindex->objects = xrealloc(eindex->objects,
>> - eindex->alloc * sizeof(struct object *));
>> - eindex->hashes = xrealloc(eindex->hashes,
>> - eindex->alloc * sizeof(uint32_t));
>> - }
>> -
>> - bitmap_pos = eindex->count;
>> - eindex->objects[eindex->count] = object;
>> - eindex->hashes[eindex->count] = pack_name_hash(name);
>> - kh_value(eindex->positions, hash_pos) = bitmap_pos;
>> - eindex->count++;
>> - } else {
>> - bitmap_pos = kh_value(eindex->positions, hash_pos);
>> - }
>> -
>> - return bitmap_pos + bitmap_git.pack->num_objects;
>> + struct ext_entry *e = xmalloc(sizeof(struct ext_entry));
>> + hashmap_entry_init(e, __kh_oid_hash(object->sha1));
>> + e->object = object;
>> + e->name_hash = pack_name_hash(name);
>> + e->nr = bitmap_git.ext_index.size;
>> + hashmap_add(&bitmap_git.ext_index, e);
>> + return e->nr + bitmap_git.pack->num_objects;
>
> One of the side effects of the current system is that the array
> effectively works as a custom allocator, and we do not pay per-entry
> malloc overhead. As I mentioned above, this is not that
> heavily-exercised a code path, so it may not matter (and you can convert
> it by explicitly using a custom allocator, at which point you could drop
> the e->nr field entirely, I'd think).
>
Realloc doesn't produce stable pointers, so we'd have to use a custom slab-allocator such as in alloc.c. However, I tend to think of malloc as a separate problem. There are quite efficient heap implementations with near-zero memory overhead even for small objects out there. I wouldn't be surprised if some day allocating many small objects will be faster and more space-efficient than re-allocating big blocks. Then you'd have all your code base littered with custom allocators that actually slow down the application...
>> @@ -584,17 +581,15 @@ static struct bitmap *find_objects(struct rev_info *revs,
>> static void show_extended_objects(struct bitmap *objects,
>> show_reachable_fn show_reach)
>> {
>> - struct eindex *eindex = &bitmap_git.ext_index;
>> - uint32_t i;
>> -
>> - for (i = 0; i < eindex->count; ++i) {
>> - struct object *obj;
>> + struct ext_entry *e;
>> + struct hashmap_iter iter;
>>
>> - if (!bitmap_get(objects, bitmap_git.pack->num_objects + i))
>> + for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
>> + e = hashmap_iter_next(&iter)) {
>> + if (!bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
>> continue;
>>
>> - obj = eindex->objects[i];
>> - show_reach(obj->sha1, obj->type, 0, eindex->hashes[i], NULL, 0);
>> + show_reach(e->object->sha1, e->object->type, 0, e->name_hash, NULL, 0);
>
> Before we were iterating in eindex order. Now we are iterating in
> hashmap order. That effects our traversal order, and ultimately our
> output for something like "rev-list". We'd want to sort on e->nr (or
> again, a custom allocator would make this go away because we would
> iterate over the array).
>
I didn't think it would matter, as the caller emits the objects from the bitmapped pack ordered by type and pack-index...well, I'll try to think of something (probably a linked list between entries?)
>> - for (i = 0; i < eindex->count; ++i) {
>> - if (eindex->objects[i]->type == type &&
>> - bitmap_get(objects, bitmap_git.pack->num_objects + i))
>> + for (e = hashmap_iter_first(&bitmap_git.ext_index, &iter); e;
>> + e = hashmap_iter_next(&iter)) {
>> + if (e->object->type == type &&
>> + bitmap_get(objects, bitmap_git.pack->num_objects + e->nr))
>> count++;
>
> Ditto here, but I do not think the order matters in this instance.
>
> -Peff
>
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 11/21] pack-objects: use bitmaps when packing objects
2013-12-07 15:47 ` Thomas Rast
@ 2013-12-21 13:15 ` Jeff King
0 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-12-21 13:15 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Sat, Dec 07, 2013 at 04:47:20PM +0100, Thomas Rast wrote:
> > +static off_t write_reused_pack(struct sha1file *f)
> > +{
> > + uint8_t buffer[8192];
>
> We usually just call this 'unsigned char'. I can see why this would be
> more portable, but git would already fall apart badly on an architecture
> where char is not 8 bits.
I think it's worth switching just for consistency with the rest of git.
Fixed.
> } + packfile_size = write_reused_pack(f);
> } + if (!packfile_size)
> } + die_errno("failed to re-use existing pack");
>
> So if you just died here, when the error happens, you could take the
> chance to tell the user _which_ syscall failed.
Yeah, agreed (and especially the fact that we may get bogus errno
values). Fixed.
> Not your fault, but sha1write() is an odd function -- it purportedly is
>
> int sha1write(struct sha1file *f, const void *buf, unsigned int count);
>
> but it can only return 0. This goes back all the way to c38138c
> (git-pack-objects: write the pack files with a SHA1 csum, 2005-06-26).
It looks like there's exactly one site that checks its return value, and
it's just to die. We should drop the return value from sha1write
entirely to make it clear that it dies on error. But that's orthogonal
to this series.
> > -static int add_object_entry(const unsigned char *sha1, enum object_type type,
> > - const char *name, int exclude)
> > +static int add_object_entry_1(const unsigned char *sha1, enum object_type type,
> > + int flags, uint32_t name_hash,
> > + struct packed_git *found_pack, off_t found_offset)
> [...]
> This function makes my head spin, and you're indenting it yet another
> level.
Yeah. In addition, the use of the "flags" here is somewhat questionable.
We use them for internal values in the call from add_object_entry to
add_object_entry_1. But we also pass the latter as a traversal callback,
meaning that what it would get in "flags" is totally different. It's not
actually a bug in the current code, since the bitmap traversal always
passes empty flags, but it's still rather confusing.
> If it's not too much work, can you split it into the three parts that it
> really is? IIUC it boils down to
>
> do we have this already?
> possibly apply 'exclude', then return
> are we coming from a call path that doesn't tell us which pack to take
> it from?
> find _all_ instances in packs
> check if any of them are local .keep packs
> if so, return
> construct a packlist entry to taste
I did this split. By itself, I was on the fence, as there are actually
some interdependencies between the three parts that make it hairy.
But then I realized that instead of making the weird relationship
between add_object_entry and add_object_entry_1, we can simply make a
new function that composes the functions differently.
So we get:
+static int add_object_entry_from_bitmap(const unsigned char *sha1,
+ enum object_type type,
+ int flags, uint32_t name_hash,
+ struct packed_git *pack, off_t offset)
+{
+ uint32_t index_pos;
+
+ if (have_duplicate_entry(sha1, 0, &index_pos))
+ return 0;
+
+ create_object_entry(sha1, type, name_hash, 0, 0, index_pos, pack, offset);
+
+ display_progress(progress_state, to_pack.nr_objects);
+ return 1;
+}
[...]
+ traverse_bitmap_commit_list(&add_object_entry_from_bitmap);
which makes much more sense.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 13/21] pack-objects: implement bitmap writing
2013-12-07 16:32 ` Thomas Rast
@ 2013-12-21 13:17 ` Jeff King
0 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-12-21 13:17 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Sat, Dec 07, 2013 at 05:32:43PM +0100, Thomas Rast wrote:
> > +pack.writebitmaps::
> > + When true, git will write a bitmap index when packing all
> > + objects to disk (e.g., as when `git repack -a` is run). This
> ^^
>
> Doesn't sound right in my ears. Remove the "as"?
It was meant to be "such as when...", but I think it is just as clear to
drop it. Will do.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 19/21] t: add basic bitmap functionality tests
2013-12-07 16:43 ` Thomas Rast
@ 2013-12-21 13:22 ` Jeff King
0 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-12-21 13:22 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Sat, Dec 07, 2013 at 05:43:29PM +0100, Thomas Rast wrote:
> One nit:
>
> > +test_expect_success JGIT 'jgit can read our bitmaps' '
> > + git clone . compat-us.git &&
> > + (
> > + cd compat-us.git &&
>
> The name suggests a bare repo, but it is a full clone. Not that it
> matters.
It was originally supposed to be a bare repo, but I had trouble
convincing jgit to actually run in a bare repo. The solution was to
switch to a non-bare one. :)
But obviously I forgot to update the name. I agree it's better to use a
more obvious name. Fixed.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
* Re: [PATCH v3 20/21] t/perf: add tests for pack bitmaps
2013-12-07 16:51 ` Thomas Rast
@ 2013-12-21 13:40 ` Jeff King
0 siblings, 0 replies; 55+ messages in thread
From: Jeff King @ 2013-12-21 13:40 UTC (permalink / raw)
To: Thomas Rast; +Cc: git, Vicent Martí
On Sat, Dec 07, 2013 at 05:51:43PM +0100, Thomas Rast wrote:
> Jeff King <peff@peff.net> writes:
>
> > +test_perf 'simulated fetch' '
> > + have=$(git rev-list HEAD --until=1.week.ago -1) &&
>
> This will give you HEAD if your GIT_PERF_LARGE_REPO hasn't seen any
> activity lately. I'd prefer something that always takes a fixed commit,
> e.g. HEAD~1000, keeping the perf test reproducible over time (not over
> changing GIT_PERF_LARGE_REPO, of course).
Good point. I just did a "git pull" on the kernel before my tests, but
of course the same test with my 3-week-stale repo would be a lot less
interesting.
I tried to figure out what is the right N for HEAD~N to represent a
week's worth of commits. It turns out the kernel is really bursty
depending on where they are in the cycle. Sometimes N is 50, and
sometimes 200. I'm picking 100. It doesn't have to be a week; the main
goal was to approximate "here's how it performs if you don't repack for
a week", so anything in the ballpark is fine.
-Peff
^ permalink raw reply [flat|nested] 55+ messages in thread
end of thread, other threads:[~2013-12-21 13:40 UTC | newest]
Thread overview: 55+ messages (download: mbox.gz follow: Atom feed
-- links below jump to the message on this page --
2013-11-14 12:41 [PATCH v3 0/21] pack bitmaps Jeff King
2013-11-14 12:42 ` [PATCH v3 01/21] sha1write: make buffer const-correct Jeff King
2013-11-14 12:42 ` [PATCH v3 02/21] revindex: Export new APIs Jeff King
2013-11-14 12:42 ` [PATCH v3 03/21] pack-objects: Refactor the packing list Jeff King
2013-11-14 12:43 ` [PATCH v3 04/21] pack-objects: factor out name_hash Jeff King
2013-11-14 12:43 ` [PATCH v3 05/21] revision: allow setting custom limiter function Jeff King
2013-11-14 12:43 ` [PATCH v3 06/21] sha1_file: export `git_open_noatime` Jeff King
2013-11-14 12:43 ` [PATCH v3 07/21] compat: add endianness helpers Jeff King
2013-11-14 12:43 ` [PATCH v3 08/21] ewah: compressed bitmap implementation Jeff King
2013-11-14 12:44 ` [PATCH v3 09/21] documentation: add documentation for the bitmap format Jeff King
2013-11-14 12:44 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Jeff King
2013-11-24 21:36 ` Thomas Rast
2013-11-25 15:04 ` [PATCH] Document khash Thomas Rast
2013-11-28 10:35 ` Jeff King
2013-11-27 9:08 ` [PATCH v3 10/21] pack-bitmap: add support for bitmap indexes Karsten Blees
2013-11-28 10:38 ` Jeff King
2013-12-03 14:40 ` Karsten Blees
2013-12-03 18:21 ` Jeff King
2013-12-07 20:52 ` Karsten Blees
2013-11-29 21:21 ` Thomas Rast
2013-12-02 16:12 ` Jeff King
2013-12-02 20:36 ` Junio C Hamano
2013-12-02 20:47 ` Jeff King
2013-12-02 21:43 ` Junio C Hamano
2013-11-14 12:45 ` [PATCH v3 11/21] pack-objects: use bitmaps when packing objects Jeff King
2013-12-07 15:47 ` Thomas Rast
2013-12-21 13:15 ` Jeff King
2013-11-14 12:45 ` [PATCH v3 12/21] rev-list: add bitmap mode to speed up object lists Jeff King
2013-12-07 16:05 ` Thomas Rast
2013-11-14 12:45 ` [PATCH v3 13/21] pack-objects: implement bitmap writing Jeff King
2013-12-07 16:32 ` Thomas Rast
2013-12-21 13:17 ` Jeff King
2013-11-14 12:45 ` [PATCH v3 14/21] repack: stop using magic number for ARRAY_SIZE(exts) Jeff King
2013-12-07 16:34 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 15/21] repack: turn exts array into array-of-struct Jeff King
2013-12-07 16:34 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 16/21] repack: handle optional files created by pack-objects Jeff King
2013-12-07 16:35 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 17/21] repack: consider bitmaps when performing repacks Jeff King
2013-12-07 16:37 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 18/21] count-objects: recognize .bitmap in garbage-checking Jeff King
2013-12-07 16:38 ` Thomas Rast
2013-11-14 12:46 ` [PATCH v3 19/21] t: add basic bitmap functionality tests Jeff King
2013-12-07 16:43 ` Thomas Rast
2013-12-21 13:22 ` Jeff King
2013-11-14 12:48 ` [PATCH v3 20/21] t/perf: add tests for pack bitmaps Jeff King
2013-12-07 16:51 ` Thomas Rast
2013-12-21 13:40 ` Jeff King
2013-11-14 12:48 ` [PATCH v3 21/21] pack-bitmap: implement optional name_hash cache Jeff King
2013-12-07 16:59 ` Thomas Rast
2013-11-14 19:19 ` [PATCH v3 0/21] pack bitmaps Ramsay Jones
2013-11-14 21:33 ` Jeff King
2013-11-14 23:09 ` Ramsay Jones
2013-11-18 21:16 ` Ramsay Jones
2013-11-16 10:28 ` Thomas Rast
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).