[FFmpeg-devel] [PATCH 3/8] avutil/mem: Add av_fast_realloc_array()

[Andreas Rheinhardt <andreas.rheinhardt@outlook.com>]

From: Andreas Rheinhardt <andres.rheinhardt@outlook.com>

This is an array-equivalent of av_fast_realloc(). Its advantages
compared to using av_fast_realloc() for allocating arrays are as
follows:

a) It performs its own overflow checks for the multiplication that is
implicit in array allocations. (And it only needs to perform these
checks (as well as the multiplication itself) in case the array needs to
be reallocated.)
b) It allows to limit the number of elements to an upper bound given
by the caller. This allows to restrict the number of allocated elements
to fit into an int and therefore makes this function usable with
counters of this type. It can also be used to avoid overflow checks in
the caller: E.g. setting it to UINT_MAX - 1 elements makes it safe to
increase the desired number of elements in steps of one. And it avoids
overallocations in situations where one already has an upper bound.
c) av_fast_realloc_array() will always allocate in multiples of array
elements; no memory is wasted with partial elements.
d) By returning an int, av_fast_realloc_array() can distinguish between
ordinary allocation failures (meriting AVERROR(ENOMEM)) and failures
because of allocation limits (by returning AVERROR(ERANGE)).
e) It is no longer possible for the user to accidentally lose the
pointer by using ptr = av_fast_realloc(ptr, ...).

Because of e) there is no need to set the number of allocated elements
to zero on failure.

av_fast_realloc() usually allocates size + size / 16 + 32 bytes if size
bytes are desired and if the already existing buffer isn't big enough.
av_fast_realloc_array() instead allocates nb + (nb + 14) / 16. Rounding
up is done in order not to reallocate in steps of one if the current
number is < 16; adding 14 instead of 15 has the effect of only
allocating one element if one element is desired. This is done with an
eye towards applications where arrays might commonly only contain one
element (as happens with the Matroska CueTrackPositions).

Which of the two functions allocates faster depends upon the size of
the elements. E.g. if the elements have a size of 32B and the desired
size is incremented in steps of one, allocations happen at
1, 3, 5, 7, 9, 11, 13, 15, 17, 20, 23, 26 ... for av_fast_realloc(),
1, 2, 4, 6, 8, 10, 12, 14, 16, 18, 21, 24 ... for
av_fast_realloc_array(). For element sizes of 96B, the numbers are
1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 13, 15, 17, 19, 21, 23, 25, 27, 30 ...
for av_fast_realloc() whereas the pattern for av_fast_realloc_array() is
unchanged.

Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt@outlook.com>
---
This patch (and some of the other patches of this patchset)
are mostly the same as the one in these threads:
https://ffmpeg.org/pipermail/ffmpeg-devel/2019-December/254836.html
https://ffmpeg.org/pipermail/ffmpeg-devel/2020-January/255182.html

 doc/APIchanges      |  3 +++
 libavutil/mem.c     | 33 +++++++++++++++++++++++++++++++++
 libavutil/mem.h     | 30 ++++++++++++++++++++++++++++++
 libavutil/version.h |  2 +-
 4 files changed, 67 insertions(+), 1 deletion(-)

diff --git a/doc/APIchanges b/doc/APIchanges
index 20b944933a..f633ae6fee 100644
--- a/doc/APIchanges
+++ b/doc/APIchanges
@@ -14,6 +14,9 @@ libavutil:     2021-04-27
 
 API changes, most recent first:
 
+2022-07-05 - xxxxxxxxxx - lavu 57.28.100 - mem.h
+  Add av_fast_realloc_array() to simplify reallocating of arrays.
+
 2022-06-12 - xxxxxxxxxx - lavf 59.25.100 - avio.h
   Add avio_vprintf(), similar to avio_printf() but allow to use it
   from within a function taking a variable argument list as input.
diff --git a/libavutil/mem.c b/libavutil/mem.c
index 18aff5291f..6e3942ae63 100644
--- a/libavutil/mem.c
+++ b/libavutil/mem.c
@@ -532,6 +532,39 @@ void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size)
     return ptr;
 }
 
+int av_fast_realloc_array(void *ptr, size_t *nb_allocated,
+                          size_t min_nb, size_t max_nb, size_t elsize)
+{
+    void *array;
+    size_t nb, max_alloc_size_bytes;
+
+    if (min_nb <= *nb_allocated)
+        return 0;
+
+    max_alloc_size_bytes = atomic_load_explicit(&max_alloc_size, memory_order_relaxed);
+    max_nb = FFMIN(max_nb, max_alloc_size_bytes / elsize);
+
+    if (min_nb > max_nb)
+        return AVERROR(ERANGE);
+
+    nb = min_nb + (min_nb + 14) / 16;
+
+    /* If min_nb is so big that the above calculation overflowed,
+     * just allocate as much as we are allowed to. */
+    nb = nb < min_nb ? max_nb : FFMIN(nb, max_nb);
+
+    memcpy(&array, ptr, sizeof(array));
+
+    array = av_realloc(array, nb * elsize);
+    if (!array)
+        return AVERROR(ENOMEM);
+
+    memcpy(ptr, &array, sizeof(array));
+    *nb_allocated = nb;
+
+    return 0;
+}
+
 static inline void fast_malloc(void *ptr, unsigned int *size, size_t min_size, int zero_realloc)
 {
     size_t max_size;
diff --git a/libavutil/mem.h b/libavutil/mem.h
index d91174196c..f040de08fc 100644
--- a/libavutil/mem.h
+++ b/libavutil/mem.h
@@ -375,11 +375,41 @@ int av_reallocp_array(void *ptr, size_t nmemb, size_t size);
  * @return `ptr` if the buffer is large enough, a pointer to newly reallocated
  *         buffer if the buffer was not large enough, or `NULL` in case of
  *         error
+ * @see av_fast_realloc_array()
  * @see av_realloc()
  * @see av_fast_malloc()
  */
 void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size);
 
+/**
+ * Reallocate the given array if it is not large enough, otherwise do nothing.
+ *
+ * If `ptr` points to `NULL`, then a new uninitialized array is allocated.
+ *
+ * @param[in,out] ptr           Pointer to `NULL` or pointer to an already
+ *                              allocated array. `*ptr` will be set to point
+ *                              to the new array on success.
+ * @param[in,out] nb_allocated  Pointer to the number of elements of the array
+ *                              `*ptr`. `*nb_allocated` is updated to the new
+ *                              number of allocated elements.
+ * @param[in]     min_nb        Desired minimal number of elements in array `*ptr`
+ * @param[in]     max_nb        Maximal number of elements to allocate.
+ * @param[in]     elsize        Size of a single element of the array.
+ *                              Must not be zero.
+ * @return 0 on success, < 0 on failure. On failure, `*ptr` is not freed and
+ *         `*ptr` as well as `*nb_allocated` are unchanged.
+ * @note `max_nb` can be used to limit allocations and make this function
+ *       usable with counters of types other than size_t. It can also be used
+ *       to avoid overflow checks in callers: E.g. setting it to `UINT_MAX - 1`
+ *       means that incrementing an unsigned int in steps of one need not be
+ *       checked for overflow.
+ * @see av_fast_realloc()
+ * @see av_realloc()
+ * @see av_fast_malloc()
+ */
+int av_fast_realloc_array(void *ptr, size_t *nb_allocated,
+                          size_t min_nb, size_t max_nb, size_t elsize);
+
 /**
  * Allocate a buffer, reusing the given one if large enough.
  *
diff --git a/libavutil/version.h b/libavutil/version.h
index 2e9e02dda8..87178e9e9a 100644
--- a/libavutil/version.h
+++ b/libavutil/version.h
@@ -79,7 +79,7 @@
  */
 
 #define LIBAVUTIL_VERSION_MAJOR  57
-#define LIBAVUTIL_VERSION_MINOR  27
+#define LIBAVUTIL_VERSION_MINOR  28
 #define LIBAVUTIL_VERSION_MICRO 100
 
 #define LIBAVUTIL_VERSION_INT   AV_VERSION_INT(LIBAVUTIL_VERSION_MAJOR, \
-- 
2.34.1

_______________________________________________
ffmpeg-devel mailing list
ffmpeg-devel@ffmpeg.org
https://ffmpeg.org/mailman/listinfo/ffmpeg-devel

To unsubscribe, visit link above, or email
ffmpeg-devel-request@ffmpeg.org with subject "unsubscribe".