mirror/ git
1
0
Fork 0
mirror of https://github.com/git/git.git synced 2024-05-07 17:26:09 +02:00
Code

Merge branch 'ps/reftable-block-iteration-optim' 

The code to iterate over reftable blocks has seen some optimization
to reduce memory allocation and deallocation.

* ps/reftable-block-iteration-optim:
  reftable/block: avoid copying block iterators on seek
  reftable/block: reuse `zstream` state on inflation
  reftable/block: open-code call to `uncompress2()`
  reftable/block: reuse uncompressed blocks
  reftable/reader: iterate to next block in place
  reftable/block: move ownership of block reader into `struct table_iter`
  reftable/block: introduce `block_reader_release()`
  reftable/block: better grouping of functions
  reftable/block: merge `block_iter_seek()` and `block_reader_seek()`
  reftable/block: rename `block_reader_start()`
This commit is contained in:
Junio C Hamano 2024-04-23 11:52:37 -07:00
parent 00e10ef10e 9da5c992dd
commit 33bbc21c92
5 changed files with 229 additions and 178 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

176
reftable/block.c
View File

@ -175,11 +175,6 @@ int block_writer_finish(struct block_writer *w)
return w->next;
}
uint8_t block_reader_type(struct block_reader *r)
{
return r->block.data[r->header_off];
}
int block_reader_init(struct block_reader *br, struct reftable_block *block,
uint32_t header_off, uint32_t table_block_size,
int hash_size)
@ -191,7 +186,8 @@ int block_reader_init(struct block_reader *br, struct reftable_block *block,
uint16_t restart_count = 0;
uint32_t restart_start = 0;
uint8_t *restart_bytes = NULL;
uint8_t *uncompressed = NULL;
reftable_block_done(&br->block);
if (!reftable_is_block_type(typ)) {
err = REFTABLE_FORMAT_ERROR;
@ -199,37 +195,57 @@ int block_reader_init(struct block_reader *br, struct reftable_block *block,
}
if (typ == BLOCK_TYPE_LOG) {
int block_header_skip = 4 + header_off;
uLongf dst_len = sz - block_header_skip; /* total size of dest
buffer. */
uLongf src_len = block->len - block_header_skip;
uint32_t block_header_skip = 4 + header_off;
uLong dst_len = sz - block_header_skip;
uLong src_len = block->len - block_header_skip;
/* Log blocks specify the *uncompressed* size in their header. */
REFTABLE_ALLOC_ARRAY(uncompressed, sz);
REFTABLE_ALLOC_GROW(br->uncompressed_data, sz,
br->uncompressed_cap);
/* Copy over the block header verbatim. It's not compressed. */
memcpy(uncompressed, block->data, block_header_skip);
memcpy(br->uncompressed_data, block->data, block_header_skip);
/* Uncompress */
if (Z_OK !=
uncompress2(uncompressed + block_header_skip, &dst_len,
block->data + block_header_skip, &src_len)) {
if (!br->zstream) {
REFTABLE_CALLOC_ARRAY(br->zstream, 1);
err = inflateInit(br->zstream);
} else {
err = inflateReset(br->zstream);
}
if (err != Z_OK) {
err = REFTABLE_ZLIB_ERROR;
goto done;
}
if (dst_len + block_header_skip != sz) {
br->zstream->next_in = block->data + block_header_skip;
br->zstream->avail_in = src_len;
br->zstream->next_out = br->uncompressed_data + block_header_skip;
br->zstream->avail_out = dst_len;
/*
* We know both input as well as output size, and we know that
* the sizes should never be bigger than `uInt_MAX` because
* blocks can at most be 16MB large. We can thus use `Z_FINISH`
* here to instruct zlib to inflate the data in one go, which
* is more efficient than using `Z_NO_FLUSH`.
*/
err = inflate(br->zstream, Z_FINISH);
if (err != Z_STREAM_END) {
err = REFTABLE_ZLIB_ERROR;
goto done;
}
err = 0;
if (br->zstream->total_out + block_header_skip != sz) {
err = REFTABLE_FORMAT_ERROR;
goto done;
}
/* We're done with the input data. */
reftable_block_done(block);
block->data = uncompressed;
uncompressed = NULL;
block->data = br->uncompressed_data;
block->len = sz;
block->source = malloc_block_source();
full_block_size = src_len + block_header_skip;
full_block_size = src_len + block_header_skip - br->zstream->avail_in;
} else if (full_block_size == 0) {
full_block_size = sz;
} else if (sz < full_block_size && sz < block->len &&
@ -257,18 +273,52 @@ int block_reader_init(struct block_reader *br, struct reftable_block *block,
br->restart_bytes = restart_bytes;
done:
reftable_free(uncompressed);
return err;
}
static uint32_t block_reader_restart_offset(struct block_reader *br, int i)
void block_reader_release(struct block_reader *br)
{
inflateEnd(br->zstream);
reftable_free(br->zstream);
reftable_free(br->uncompressed_data);
reftable_block_done(&br->block);
}
uint8_t block_reader_type(const struct block_reader *r)
{
return r->block.data[r->header_off];
}
int block_reader_first_key(const struct block_reader *br, struct strbuf *key)
{
int off = br->header_off + 4, n;
struct string_view in = {
.buf = br->block.data + off,
.len = br->block_len - off,
};
uint8_t extra = 0;
strbuf_reset(key);
n = reftable_decode_key(key, &extra, in);
if (n < 0)
return n;
if (!key->len)
return REFTABLE_FORMAT_ERROR;
return 0;
}
static uint32_t block_reader_restart_offset(const struct block_reader *br, int i)
{
return get_be24(br->restart_bytes + 3 * i);
}
void block_reader_start(struct block_reader *br, struct block_iter *it)
void block_iter_seek_start(struct block_iter *it, const struct block_reader *br)
{
it->br = br;
it->block = br->block.data;
it->block_len = br->block_len;
it->hash_size = br->hash_size;
strbuf_reset(&it->last_key);
it->next_off = br->header_off + 4;
}
@ -276,7 +326,7 @@ void block_reader_start(struct block_reader *br, struct block_iter *it)
struct restart_needle_less_args {
int error;
struct strbuf needle;
struct block_reader *reader;
const struct block_reader *reader;
};
static int restart_needle_less(size_t idx, void *_args)
@ -315,25 +365,17 @@ static int restart_needle_less(size_t idx, void *_args)
return args->needle.len < suffix_len;
}
void block_iter_copy_from(struct block_iter *dest, struct block_iter *src)
{
dest->br = src->br;
dest->next_off = src->next_off;
strbuf_reset(&dest->last_key);
strbuf_addbuf(&dest->last_key, &src->last_key);
}
int block_iter_next(struct block_iter *it, struct reftable_record *rec)
{
struct string_view in = {
.buf = it->br->block.data + it->next_off,
.len = it->br->block_len - it->next_off,
.buf = (unsigned char *) it->block + it->next_off,
.len = it->block_len - it->next_off,
};
struct string_view start = in;
uint8_t extra = 0;
int n = 0;
if (it->next_off >= it->br->block_len)
if (it->next_off >= it->block_len)
return 1;
n = reftable_decode_key(&it->last_key, &extra, in);
@ -343,7 +385,7 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
return REFTABLE_FORMAT_ERROR;
string_view_consume(&in, n);
n = reftable_record_decode(rec, it->last_key, extra, in, it->br->hash_size,
n = reftable_record_decode(rec, it->last_key, extra, in, it->hash_size,
&it->scratch);
if (n < 0)
return -1;
@ -353,29 +395,13 @@ int block_iter_next(struct block_iter *it, struct reftable_record *rec)
return 0;
}
int block_reader_first_key(struct block_reader *br, struct strbuf *key)
void block_iter_reset(struct block_iter *it)
{
int off = br->header_off + 4, n;
struct string_view in = {
.buf = br->block.data + off,
.len = br->block_len - off,
};
uint8_t extra = 0;
strbuf_reset(key);
n = reftable_decode_key(key, &extra, in);
if (n < 0)
return n;
if (!key->len)
return REFTABLE_FORMAT_ERROR;
return 0;
}
int block_iter_seek(struct block_iter *it, struct strbuf *want)
{
return block_reader_seek(it->br, it, want);
strbuf_reset(&it->last_key);
it->next_off = 0;
it->block = NULL;
it->block_len = 0;
it->hash_size = 0;
}
void block_iter_close(struct block_iter *it)
@ -384,14 +410,13 @@ void block_iter_close(struct block_iter *it)
strbuf_release(&it->scratch);
}
int block_reader_seek(struct block_reader *br, struct block_iter *it,
struct strbuf *want)
int block_iter_seek_key(struct block_iter *it, const struct block_reader *br,
struct strbuf *want)
{
struct restart_needle_less_args args = {
.needle = *want,
.reader = br,
};
struct block_iter next = BLOCK_ITER_INIT;
struct reftable_record rec;
int err = 0;
size_t i;
@ -436,7 +461,9 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
it->next_off = block_reader_restart_offset(br, i - 1);
else
it->next_off = br->header_off + 4;
it->br = br;
it->block = br->block.data;
it->block_len = br->block_len;
it->hash_size = br->hash_size;
reftable_record_init(&rec, block_reader_type(br));
@ -448,11 +475,13 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
* far and then back up.
*/
while (1) {
block_iter_copy_from(&next, it);
err = block_iter_next(&next, &rec);
size_t prev_off = it->next_off;
err = block_iter_next(it, &rec);
if (err < 0)
goto done;
if (err > 0) {
it->next_off = prev_off;
err = 0;
goto done;
}
@ -463,18 +492,23 @@ int block_reader_seek(struct block_reader *br, struct block_iter *it,
* record does not exist in the block and can thus abort early.
* In case it is equal to the sought-after key we have found
* the desired record.
*
* Note that we store the next record's key record directly in
* `last_key` without restoring the key of the preceding record
* in case we need to go one record back. This is safe to do as
* `block_iter_next()` would return the ref whose key is equal
* to `last_key` now, and naturally all keys share a prefix
* with themselves.
*/
reftable_record_key(&rec, &it->last_key);
if (strbuf_cmp(&it->last_key, want) >= 0)
if (strbuf_cmp(&it->last_key, want) >= 0) {
it->next_off = prev_off;
goto done;
block_iter_copy_from(it, &next);
}
}
done:
block_iter_close(&next);
reftable_record_release(&rec);
return err;
}

47
reftable/block.h
View File

@ -56,6 +56,8 @@ int block_writer_finish(struct block_writer *w);
/* clears out internally allocated block_writer members. */
void block_writer_release(struct block_writer *bw);
struct z_stream;
/* Read a block. */
struct block_reader {
/* offset of the block header; nonzero for the first block in a
@ -66,6 +68,11 @@ struct block_reader {
struct reftable_block block;
int hash_size;
/* Uncompressed data for log entries. */
z_stream *zstream;
unsigned char *uncompressed_data;
size_t uncompressed_cap;
/* size of the data, excluding restart data. */
uint32_t block_len;
uint8_t *restart_bytes;
@ -76,11 +83,26 @@ struct block_reader {
uint32_t full_block_size;
};
/* initializes a block reader. */
int block_reader_init(struct block_reader *br, struct reftable_block *bl,
uint32_t header_off, uint32_t table_block_size,
int hash_size);
void block_reader_release(struct block_reader *br);
/* Returns the block type (eg. 'r' for refs) */
uint8_t block_reader_type(const struct block_reader *r);
/* Decodes the first key in the block */
int block_reader_first_key(const struct block_reader *br, struct strbuf *key);
/* Iterate over entries in a block */
struct block_iter {
/* offset within the block of the next entry to read. */
uint32_t next_off;
struct block_reader *br;
const unsigned char *block;
size_t block_len;
int hash_size;
/* key for last entry we read. */
struct strbuf last_key;
@ -92,31 +114,18 @@ struct block_iter {
.scratch = STRBUF_INIT, \
}
/* initializes a block reader. */
int block_reader_init(struct block_reader *br, struct reftable_block *bl,
uint32_t header_off, uint32_t table_block_size,
int hash_size);
/* Position `it` at start of the block */
void block_reader_start(struct block_reader *br, struct block_iter *it);
void block_iter_seek_start(struct block_iter *it, const struct block_reader *br);
/* Position `it` to the `want` key in the block */
int block_reader_seek(struct block_reader *br, struct block_iter *it,
struct strbuf *want);
/* Returns the block type (eg. 'r' for refs) */
uint8_t block_reader_type(struct block_reader *r);
/* Decodes the first key in the block */
int block_reader_first_key(struct block_reader *br, struct strbuf *key);
void block_iter_copy_from(struct block_iter *dest, struct block_iter *src);
int block_iter_seek_key(struct block_iter *it, const struct block_reader *br,
struct strbuf *want);
/* return < 0 for error, 0 for OK, > 0 for EOF. */
int block_iter_next(struct block_iter *it, struct reftable_record *rec);
/* Seek to `want` with in the block pointed to by `it` */
int block_iter_seek(struct block_iter *it, struct strbuf *want);
/* Reset the block iterator to pristine state without releasing its memory. */
void block_iter_reset(struct block_iter *it);
/* deallocate memory for `it`. The block reader and its block is left intact. */
void block_iter_close(struct block_iter *it);

6
reftable/block_test.c
View File

@ -69,7 +69,7 @@ static void test_block_read_write(void)
block_reader_init(&br, &block, header_off, block_size, GIT_SHA1_RAWSZ);
block_reader_start(&br, &it);
block_iter_seek_start(&it, &br);
while (1) {
int r = block_iter_next(&it, &rec);
@ -89,7 +89,7 @@ static void test_block_read_write(void)
strbuf_reset(&want);
strbuf_addstr(&want, names[i]);
n = block_reader_seek(&br, &it, &want);
n = block_iter_seek_key(&it, &br, &want);
EXPECT(n == 0);
n = block_iter_next(&it, &rec);
@ -98,7 +98,7 @@ static void test_block_read_write(void)
EXPECT_STREQ(names[i], rec.u.ref.refname);
want.len--;
n = block_reader_seek(&br, &it, &want);
n = block_iter_seek_key(&it, &br, &want);
EXPECT(n == 0);
n = block_iter_next(&it, &rec);

2
reftable/iter.c
View File

@ -115,7 +115,7 @@ static int indexed_table_ref_iter_next_block(struct indexed_table_ref_iter *it)
/* indexed block does not exist. */
return REFTABLE_FORMAT_ERROR;
}
block_reader_start(&it->block_reader, &it->cur);
block_iter_seek_start(&it->cur, &it->block_reader);
return 0;
}

176
reftable/reader.c
View File

@ -220,6 +220,7 @@ struct table_iter {
struct reftable_reader *r;
uint8_t typ;
uint64_t block_off;
struct block_reader br;
struct block_iter bi;
int is_finished;
};
@ -227,16 +228,6 @@ struct table_iter {
.bi = BLOCK_ITER_INIT \
}
static void table_iter_copy_from(struct table_iter *dest,
struct table_iter *src)
{
dest->r = src->r;
dest->typ = src->typ;
dest->block_off = src->block_off;
dest->is_finished = src->is_finished;
block_iter_copy_from(&dest->bi, &src->bi);
}
static int table_iter_next_in_block(struct table_iter *ti,
struct reftable_record *rec)
{
@ -250,14 +241,8 @@ static int table_iter_next_in_block(struct table_iter *ti,
static void table_iter_block_done(struct table_iter *ti)
{
if (!ti->bi.br) {
return;
}
reftable_block_done(&ti->bi.br->block);
FREE_AND_NULL(ti->bi.br);
ti->bi.last_key.len = 0;
ti->bi.next_off = 0;
block_reader_release(&ti->br);
block_iter_reset(&ti->bi);
}
static int32_t extract_block_size(uint8_t *data, uint8_t *typ, uint64_t off,
@ -321,32 +306,27 @@ int reader_init_block_reader(struct reftable_reader *r, struct block_reader *br,
return err;
}
static int table_iter_next_block(struct table_iter *dest,
struct table_iter *src)
static void table_iter_close(struct table_iter *ti)
{
uint64_t next_block_off = src->block_off + src->bi.br->full_block_size;
struct block_reader br = { 0 };
int err = 0;
table_iter_block_done(ti);
block_iter_close(&ti->bi);
}
dest->r = src->r;
dest->typ = src->typ;
dest->block_off = next_block_off;
static int table_iter_next_block(struct table_iter *ti)
{
uint64_t next_block_off = ti->block_off + ti->br.full_block_size;
int err;
err = reader_init_block_reader(src->r, &br, next_block_off, src->typ);
if (err > 0) {
dest->is_finished = 1;
return 1;
}
if (err != 0)
err = reader_init_block_reader(ti->r, &ti->br, next_block_off, ti->typ);
if (err > 0)
ti->is_finished = 1;
if (err)
return err;
else {
struct block_reader *brp =
reftable_malloc(sizeof(struct block_reader));
*brp = br;
dest->is_finished = 0;
block_reader_start(brp, &dest->bi);
}
ti->block_off = next_block_off;
ti->is_finished = 0;
block_iter_seek_start(&ti->bi, &ti->br);
return 0;
}
@ -356,7 +336,6 @@ static int table_iter_next(struct table_iter *ti, struct reftable_record *rec)
return REFTABLE_API_ERROR;
while (1) {
struct table_iter next = TABLE_ITER_INIT;
int err;
if (ti->is_finished)
@ -376,15 +355,11 @@ static int table_iter_next(struct table_iter *ti, struct reftable_record *rec)
* table and retry. If there are no more blocks then the
* iterator is drained.
*/
err = table_iter_next_block(&next, ti);
table_iter_block_done(ti);
err = table_iter_next_block(ti);
if (err) {
ti->is_finished = 1;
return err;
}
table_iter_copy_from(ti, &next);
block_iter_close(&next.bi);
}
}
@ -393,16 +368,14 @@ static int table_iter_next_void(void *ti, struct reftable_record *rec)
return table_iter_next(ti, rec);
}
static void table_iter_close(void *p)
static void table_iter_close_void(void *ti)
{
struct table_iter *ti = p;
table_iter_block_done(ti);
block_iter_close(&ti->bi);
table_iter_close(ti);
}
static struct reftable_iterator_vtable table_iter_vtable = {
.next = &table_iter_next_void,
.close = &table_iter_close,
.close = &table_iter_close_void,
};
static void iterator_from_table_iter(struct reftable_iterator *it,
@ -417,19 +390,16 @@ static int reader_table_iter_at(struct reftable_reader *r,
struct table_iter *ti, uint64_t off,
uint8_t typ)
{
struct block_reader br = { 0 };
struct block_reader *brp = NULL;
int err;
int err = reader_init_block_reader(r, &br, off, typ);
err = reader_init_block_reader(r, &ti->br, off, typ);
if (err != 0)
return err;
brp = reftable_malloc(sizeof(struct block_reader));
*brp = br;
ti->r = r;
ti->typ = block_reader_type(brp);
ti->typ = block_reader_type(&ti->br);
ti->block_off = off;
block_reader_start(brp, &ti->bi);
block_iter_seek_start(&ti->bi, &ti->br);
return 0;
}
@ -454,23 +424,52 @@ static int reader_seek_linear(struct table_iter *ti,
{
struct strbuf want_key = STRBUF_INIT;
struct strbuf got_key = STRBUF_INIT;
struct table_iter next = TABLE_ITER_INIT;
struct reftable_record rec;
int err = -1;
reftable_record_init(&rec, reftable_record_type(want));
reftable_record_key(want, &want_key);
/*
* First we need to locate the block that must contain our record. To
* do so we scan through blocks linearly until we find the first block
* whose first key is bigger than our wanted key. Once we have found
* that block we know that the key must be contained in the preceding
* block.
*
* This algorithm is somewhat unfortunate because it means that we
* always have to seek one block too far and then back up. But as we
* can only decode the _first_ key of a block but not its _last_ key we
* have no other way to do this.
*/
while (1) {
err = table_iter_next_block(&next, ti);
struct table_iter next = *ti;
/*
* We must be careful to not modify underlying data of `ti`
* because we may find that `next` does not contain our desired
* block, but that `ti` does. In that case, we would discard
* `next` and continue with `ti`.
*
* This also means that we cannot reuse allocated memory for
* `next` here. While it would be great if we could, it should
* in practice not be too bad given that we should only ever
* end up doing linear seeks with at most three blocks. As soon
* as we have more than three blocks we would have an index, so
* we would not do a linear search there anymore.
*/
memset(&next.br.block, 0, sizeof(next.br.block));
next.br.zstream = NULL;
next.br.uncompressed_data = NULL;
next.br.uncompressed_cap = 0;
err = table_iter_next_block(&next);
if (err < 0)
goto done;
if (err > 0) {
if (err > 0)
break;
}
err = block_reader_first_key(next.bi.br, &got_key);
err = block_reader_first_key(&next.br, &got_key);
if (err < 0)
goto done;
@ -480,16 +479,20 @@ static int reader_seek_linear(struct table_iter *ti,
}
table_iter_block_done(ti);
table_iter_copy_from(ti, &next);
*ti = next;
}
err = block_iter_seek(&ti->bi, &want_key);
/*
* We have located the block that must contain our record, so we seek
* the wanted key inside of it. If the block does not contain our key
* we know that the corresponding record does not exist.
*/
err = block_iter_seek_key(&ti->bi, &ti->br, &want_key);
if (err < 0)
goto done;
err = 0;
done:
block_iter_close(&next.bi);
reftable_record_release(&rec);
strbuf_release(&want_key);
strbuf_release(&got_key);
@ -508,6 +511,7 @@ static int reader_seek_indexed(struct reftable_reader *r,
.u.idx = { .last_key = STRBUF_INIT },
};
struct table_iter index_iter = TABLE_ITER_INIT;
struct table_iter empty = TABLE_ITER_INIT;
struct table_iter next = TABLE_ITER_INIT;
int err = 0;
@ -549,7 +553,6 @@ static int reader_seek_indexed(struct reftable_reader *r,
* not exist.
*/
err = table_iter_next(&index_iter, &index_result);
table_iter_block_done(&index_iter);
if (err != 0)
goto done;
@ -558,7 +561,7 @@ static int reader_seek_indexed(struct reftable_reader *r,
if (err != 0)
goto done;
err = block_iter_seek(&next.bi, &want_index.u.idx.last_key);
err = block_iter_seek_key(&next.bi, &next.br, &want_index.u.idx.last_key);
if (err < 0)
goto done;
@ -572,18 +575,20 @@ static int reader_seek_indexed(struct reftable_reader *r,
break;
}
table_iter_copy_from(&index_iter, &next);
table_iter_close(&index_iter);
index_iter = next;
next = empty;
}
if (err == 0) {
struct table_iter empty = TABLE_ITER_INIT;
struct table_iter *malloced = reftable_calloc(1, sizeof(*malloced));
*malloced = empty;
table_iter_copy_from(malloced, &next);
*malloced = next;
next = empty;
iterator_from_table_iter(it, malloced);
}
done:
block_iter_close(&next.bi);
table_iter_close(&next);
table_iter_close(&index_iter);
reftable_record_release(&want_index);
reftable_record_release(&index_result);
@ -597,25 +602,28 @@ static int reader_seek_internal(struct reftable_reader *r,
struct reftable_reader_offsets *offs =
reader_offsets_for(r, reftable_record_type(rec));
uint64_t idx = offs->index_offset;
struct table_iter ti = TABLE_ITER_INIT;
int err = 0;
struct table_iter ti = TABLE_ITER_INIT, *p;
int err;
if (idx > 0)
return reader_seek_indexed(r, it, rec);
err = reader_start(r, &ti, reftable_record_type(rec), 0);
if (err < 0)
return err;
goto out;
err = reader_seek_linear(&ti, rec);
if (err < 0)
return err;
else {
struct table_iter *p =
reftable_malloc(sizeof(struct table_iter));
*p = ti;
iterator_from_table_iter(it, p);
}
goto out;
return 0;
REFTABLE_ALLOC_ARRAY(p, 1);
*p = ti;
iterator_from_table_iter(it, p);
out:
if (err)
table_iter_close(&ti);
return err;
}
static int reader_seek(struct reftable_reader *r, struct reftable_iterator *it,