git/list[1] front-page[2] threads[3] people[4] search[5] about
 

[PATCH v2 1/9] reftable: introduce macros to grow arrays

From
Patrick Steinhardt <ps@pks.im>
Date
Feb 1, 2024, 07:32 UTC
Message-ID
<12bd721ddff7020eb9e9ebd4e797d50193250bc0.1706772591.git.ps@pks.im>
In-Reply-To
<cover.1706772591.git.ps@pks.im>

Throughout the reftable library we have many cases where we need to grow arrays. In order to avoid too many reallocations, we roughly double the capacity of the array on each iteration. The resulting code pattern is duplicated across many sites.

We have similar patterns in our main codebase, which is why we have eventually introduced an `ALLOC_GROW()` macro to abstract it away and avoid some code duplication. We cannot easily reuse this macro here though because `ALLOC_GROW()` uses `REALLOC_ARRAY()`, which in turn will call realloc(3P) to grow the array. The reftable code is structured as a library though (even if the boundaries are fuzzy), and one property this brings with it is that it is possible to plug in your own allocators. So instead of using realloc(3P), we need to use `reftable_realloc()` that knows to use the user-provided implementation.

So let's introduce two new macros `REFTABLE_REALLOC_ARRAY()` and `REFTABLE_ALLOC_GROW()` that mirror what we do in our main codebase, with two modifications:

  - They use `reftable_realloc()`, as explained above.
  - They use a different growth factor of `2 * cap + 1` instead of `(cap
    + 16) * 3 / 2`.

The second change is because we know a bit more about the allocation patterns in the reftable library. In most cases, we end up only having a handful of items in the array and don't end up growing them. The initial capacity that our normal growth factor uses (which is 24) would thus end up over-allocating in a lot of code paths. This effect is measurable:

  - Before change:
      HEAP SUMMARY:
          in use at exit: 671,983 bytes in 152 blocks
        total heap usage: 3,843,446 allocs, 3,843,294 frees, 223,761,402 bytes allocated
  - After change with a growth factor of `(2 * alloc + 1)`:
      HEAP SUMMARY:
          in use at exit: 671,983 bytes in 152 blocks
        total heap usage: 3,843,446 allocs, 3,843,294 frees, 223,761,410 bytes allocated
  - After change with a growth factor of `(alloc + 16)* 2 / 3`:
      HEAP SUMMARY:
          in use at exit: 671,983 bytes in 152 blocks
        total heap usage: 3,833,673 allocs, 3,833,521 frees, 4,728,251,742 bytes allocated

While the total heap usage is roughly the same, we do end up allocating significantly more bytes with our usual growth factor (in fact, roughly 21 times as many).

Convert the reftable library to use these new macros.
Signed-off-by: Patrick Steinhardt <ps@pks.im>
---
 reftable/basics.c      |  8 ++------
 reftable/basics.h      | 11 +++++++++++
 reftable/block.c       |  7 +------
 reftable/merged_test.c | 20 ++++++--------------
 reftable/pq.c          |  8 ++------
 reftable/stack.c       | 29 ++++++++++++-----------------
 reftable/writer.c      | 14 ++------------
 7 files changed, 36 insertions(+), 61 deletions(-)
diff --git a/reftable/basics.c b/reftable/basics.c
index f761e48028..af9004cec2 100644
--- a/reftable/basics.c
+++ b/reftable/basics.c
@@ -89,17 +89,13 @@ void parse_names(char *buf, int size, char ***namesp)
 			next = end;
 		}
 		if (p < next) {
-			if (names_len == names_cap) {
-				names_cap = 2 * names_cap + 1;
-				names = reftable_realloc(
-					names, names_cap * sizeof(*names));
-			}
+			REFTABLE_ALLOC_GROW(names, names_len + 1, names_cap);
 			names[names_len++] = xstrdup(p);
 		}
 		p = next + 1;
 	}
 
-	names = reftable_realloc(names, (names_len + 1) * sizeof(*names));
+	REFTABLE_REALLOC_ARRAY(names, names_len + 1);
 	names[names_len] = NULL;
 	*namesp = names;
 }
diff --git a/reftable/basics.h b/reftable/basics.h
index 096b36862b..2f855cd724 100644
--- a/reftable/basics.h
+++ b/reftable/basics.h
@@ -53,6 +53,17 @@ void *reftable_realloc(void *p, size_t sz);
 void reftable_free(void *p);
 void *reftable_calloc(size_t sz);
 
+#define REFTABLE_REALLOC_ARRAY(x, alloc) (x) = reftable_realloc((x), st_mult(sizeof(*(x)), (alloc)))
+#define REFTABLE_ALLOC_GROW(x, nr, alloc) \
+	do { \
+		if ((nr) > alloc) { \
+			alloc = 2 * (alloc) + 1; \
+			if (alloc < (nr)) \
+				alloc = (nr); \
+			REFTABLE_REALLOC_ARRAY(x, alloc); \
+		} \
+	} while (0)
+
 /* Find the longest shared prefix size of `a` and `b` */
 struct strbuf;
 int common_prefix_size(struct strbuf *a, struct strbuf *b);
diff --git a/reftable/block.c b/reftable/block.c
index 1df3d8a0f0..6952d0facf 100644
--- a/reftable/block.c
+++ b/reftable/block.c
@@ -51,12 +51,7 @@ static int block_writer_register_restart(struct block_writer *w, int n,
 	if (2 + 3 * rlen + n > w->block_size - w->next)
 		return -1;
 	if (is_restart) {
-		if (w->restart_len == w->restart_cap) {
-			w->restart_cap = w->restart_cap * 2 + 1;
-			w->restarts = reftable_realloc(
-				w->restarts, sizeof(uint32_t) * w->restart_cap);
-		}
-
+		REFTABLE_ALLOC_GROW(w->restarts, w->restart_len + 1, w->restart_cap);
 		w->restarts[w->restart_len++] = w->next;
 	}
 
diff --git a/reftable/merged_test.c b/reftable/merged_test.c
index 46908f738f..e05351e035 100644
--- a/reftable/merged_test.c
+++ b/reftable/merged_test.c
@@ -231,14 +231,10 @@ static void test_merged(void)
 	while (len < 100) { /* cap loops/recursion. */
 		struct reftable_ref_record ref = { NULL };
 		int err = reftable_iterator_next_ref(&it, &ref);
-		if (err > 0) {
+		if (err > 0)
 			break;
-		}
-		if (len == cap) {
-			cap = 2 * cap + 1;
-			out = reftable_realloc(
-				out, sizeof(struct reftable_ref_record) * cap);
-		}
+
+		REFTABLE_ALLOC_GROW(out, len + 1, cap);
 		out[len++] = ref;
 	}
 	reftable_iterator_destroy(&it);
@@ -368,14 +364,10 @@ static void test_merged_logs(void)
 	while (len < 100) { /* cap loops/recursion. */
 		struct reftable_log_record log = { NULL };
 		int err = reftable_iterator_next_log(&it, &log);
-		if (err > 0) {
+		if (err > 0)
 			break;
-		}
-		if (len == cap) {
-			cap = 2 * cap + 1;
-			out = reftable_realloc(
-				out, sizeof(struct reftable_log_record) * cap);
-		}
+
+		REFTABLE_ALLOC_GROW(out, len + 1, cap);
 		out[len++] = log;
 	}
 	reftable_iterator_destroy(&it);
diff --git a/reftable/pq.c b/reftable/pq.c
index dcefeb793a..2461daf5ff 100644
--- a/reftable/pq.c
+++ b/reftable/pq.c
@@ -75,13 +75,9 @@ void merged_iter_pqueue_add(struct merged_iter_pqueue *pq, const struct pq_entry
 {
 	int i = 0;
 
-	if (pq->len == pq->cap) {
-		pq->cap = 2 * pq->cap + 1;
-		pq->heap = reftable_realloc(pq->heap,
-					    pq->cap * sizeof(struct pq_entry));
-	}
-
+	REFTABLE_ALLOC_GROW(pq->heap, pq->len + 1, pq->cap);
 	pq->heap[pq->len++] = *e;
+
 	i = pq->len - 1;
 	while (i > 0) {
 		int j = (i - 1) / 2;
diff --git a/reftable/stack.c b/reftable/stack.c
index bf3869ce70..1dfab99e96 100644
--- a/reftable/stack.c
+++ b/reftable/stack.c
@@ -551,7 +551,7 @@ struct reftable_addition {
 	struct reftable_stack *stack;
 
 	char **new_tables;
-	int new_tables_len;
+	size_t new_tables_len, new_tables_cap;
 	uint64_t next_update_index;
 };
 
@@ -602,8 +602,9 @@ static int reftable_stack_init_addition(struct reftable_addition *add,
 
 static void reftable_addition_close(struct reftable_addition *add)
 {
-	int i = 0;
 	struct strbuf nm = STRBUF_INIT;
+	size_t i;
+
 	for (i = 0; i < add->new_tables_len; i++) {
 		stack_filename(&nm, add->stack, add->new_tables[i]);
 		unlink(nm.buf);
@@ -613,6 +614,7 @@ static void reftable_addition_close(struct reftable_addition *add)
 	reftable_free(add->new_tables);
 	add->new_tables = NULL;
 	add->new_tables_len = 0;
+	add->new_tables_cap = 0;
 
 	delete_tempfile(&add->lock_file);
 	strbuf_release(&nm);
@@ -631,8 +633,8 @@ int reftable_addition_commit(struct reftable_addition *add)
 {
 	struct strbuf table_list = STRBUF_INIT;
 	int lock_file_fd = get_tempfile_fd(add->lock_file);
-	int i = 0;
 	int err = 0;
+	size_t i;
 
 	if (add->new_tables_len == 0)
 		goto done;
@@ -660,12 +662,12 @@ int reftable_addition_commit(struct reftable_addition *add)
 	}
 
 	/* success, no more state to clean up. */
-	for (i = 0; i < add->new_tables_len; i++) {
+	for (i = 0; i < add->new_tables_len; i++)
 		reftable_free(add->new_tables[i]);
-	}
 	reftable_free(add->new_tables);
 	add->new_tables = NULL;
 	add->new_tables_len = 0;
+	add->new_tables_cap = 0;
 
 	err = reftable_stack_reload_maybe_reuse(add->stack, 1);
 	if (err)
@@ -792,11 +794,9 @@ int reftable_addition_add(struct reftable_addition *add,
 		goto done;
 	}
 
-	add->new_tables = reftable_realloc(add->new_tables,
-					   sizeof(*add->new_tables) *
-						   (add->new_tables_len + 1));
-	add->new_tables[add->new_tables_len] = strbuf_detach(&next_name, NULL);
-	add->new_tables_len++;
+	REFTABLE_ALLOC_GROW(add->new_tables, add->new_tables_len + 1,
+			    add->new_tables_cap);
+	add->new_tables[add->new_tables_len++] = strbuf_detach(&next_name, NULL);
 done:
 	if (tab_fd > 0) {
 		close(tab_fd);
@@ -1367,17 +1367,12 @@ static int stack_check_addition(struct reftable_stack *st,
 	while (1) {
 		struct reftable_ref_record ref = { NULL };
 		err = reftable_iterator_next_ref(&it, &ref);
-		if (err > 0) {
+		if (err > 0)
 			break;
-		}
 		if (err < 0)
 			goto done;
 
-		if (len >= cap) {
-			cap = 2 * cap + 1;
-			refs = reftable_realloc(refs, cap * sizeof(refs[0]));
-		}
-
+		REFTABLE_ALLOC_GROW(refs, len + 1, cap);
 		refs[len++] = ref;
 	}
 
diff --git a/reftable/writer.c b/reftable/writer.c
index ee4590e20f..4483bb21c3 100644
--- a/reftable/writer.c
+++ b/reftable/writer.c
@@ -200,12 +200,7 @@ static void writer_index_hash(struct reftable_writer *w, struct strbuf *hash)
 		return;
 	}
 
-	if (key->offset_len == key->offset_cap) {
-		key->offset_cap = 2 * key->offset_cap + 1;
-		key->offsets = reftable_realloc(
-			key->offsets, sizeof(uint64_t) * key->offset_cap);
-	}
-
+	REFTABLE_ALLOC_GROW(key->offsets, key->offset_len + 1, key->offset_cap);
 	key->offsets[key->offset_len++] = off;
 }
 
@@ -674,12 +669,7 @@ static int writer_flush_nonempty_block(struct reftable_writer *w)
 	if (err < 0)
 		return err;
 
-	if (w->index_cap == w->index_len) {
-		w->index_cap = 2 * w->index_cap + 1;
-		w->index = reftable_realloc(
-			w->index,
-			sizeof(struct reftable_index_record) * w->index_cap);
-	}
+	REFTABLE_ALLOC_GROW(w->index, w->index_len + 1, w->index_cap);
 
 	ir.offset = w->next;
 	strbuf_reset(&ir.last_key);

-- 
2.43.GIT
Previous: Patrick SteinhardtNext: Patrick Steinhardt
Message 17 of 40 in “reftable: code style improvements”
  1. 0/9 reftable: code style improvementsPatrick Steinhardt, Jan 31, 2024
  2. 1/9 reftable: introduce macros to grow arraysPatrick Steinhardt, Jan 31, 2024
  3. Eric SunshineJan 31, 2024
  4. Junio C HamanoJan 31, 2024
  5. Patrick SteinhardtFeb 1, 2024
  6. Junio C HamanoFeb 1, 2024
  7. 2/9 reftable: introduce macros to allocate arraysPatrick Steinhardt, Jan 31, 2024
  8. 3/9 reftable/stack: fix parameter validation when compacting rangePatrick Steinhardt, Jan 31, 2024
  9. 4/9 reftable/stack: index segments with `size_t`Patrick Steinhardt, Jan 31, 2024
  10. 5/9 reftable/stack: use `size_t` to track stack slices during compactionPatrick Steinhardt, Jan 31, 2024
  11. 6/9 reftable/stack: use `size_t` to track stack lengthPatrick Steinhardt, Jan 31, 2024
  12. 7/9 reftable/merged: refactor seeking of recordsPatrick Steinhardt, Jan 31, 2024
  13. Eric SunshineJan 31, 2024
  14. 8/9 reftable/merged: refactor initialization of iteratorsPatrick Steinhardt, Jan 31, 2024
  15. 9/9 reftable/record: improve semantics when initializing recordsPatrick Steinhardt, Jan 31, 2024
  16. 0/9 reftable: code style improvementsPatrick Steinhardt, Feb 1, 2024
  17. 1/9 reftable: introduce macros to grow arraysPatrick Steinhardt, Feb 1, 2024
  18. 2/9 reftable: introduce macros to allocate arraysPatrick Steinhardt, Feb 1, 2024
  19. Karthik NayakFeb 5, 2024
  20. Patrick SteinhardtFeb 6, 2024
  21. 3/9 reftable/stack: fix parameter validation when compacting rangePatrick Steinhardt, Feb 1, 2024
  22. Toon ClaesFeb 1, 2024
  23. Patrick SteinhardtFeb 2, 2024
  24. 4/9 reftable/stack: index segments with `size_t`Patrick Steinhardt, Feb 1, 2024
  25. 5/9 reftable/stack: use `size_t` to track stack slices during compactionPatrick Steinhardt, Feb 1, 2024
  26. 6/9 reftable/stack: use `size_t` to track stack lengthPatrick Steinhardt, Feb 1, 2024
  27. 7/9 reftable/merged: refactor seeking of recordsPatrick Steinhardt, Feb 1, 2024
  28. 8/9 reftable/merged: refactor initialization of iteratorsPatrick Steinhardt, Feb 1, 2024
  29. 9/9 reftable/record: improve semantics when initializing recordsPatrick Steinhardt, Feb 1, 2024
  30. 0/9 reftable: code style improvementsPatrick Steinhardt, Feb 6, 2024
  31. 1/9 reftable: introduce macros to grow arraysPatrick Steinhardt, Feb 6, 2024
  32. 2/9 reftable: introduce macros to allocate arraysPatrick Steinhardt, Feb 6, 2024
  33. 3/9 reftable/stack: fix parameter validation when compacting rangePatrick Steinhardt, Feb 6, 2024
  34. 4/9 reftable/stack: index segments with `size_t`Patrick Steinhardt, Feb 6, 2024
  35. 5/9 reftable/stack: use `size_t` to track stack slices during compactionPatrick Steinhardt, Feb 6, 2024
  36. 6/9 reftable/stack: use `size_t` to track stack lengthPatrick Steinhardt, Feb 6, 2024
  37. 7/9 reftable/merged: refactor seeking of recordsPatrick Steinhardt, Feb 6, 2024
  38. 8/9 reftable/merged: refactor initialization of iteratorsPatrick Steinhardt, Feb 6, 2024
  39. 9/9 reftable/record: improve semantics when initializing recordsPatrick Steinhardt, Feb 6, 2024
  40. Karthik NayakFeb 6, 2024

Read the whole thread, see it on lore, or plain text.

$ cat FOOTERMessages come from the public archive at lore.kernel.org/git, fetched every hour. The front page is chosen and written each morning by an AI editor and can be wrong; the threads themselves are the record. About and API. For agents: an MCP server at https://gitlist.dev/mcp, and any thread, story or person page as Markdown by adding .md to its URL (or sending Accept: text/markdown). Details in /llms.txt.