diff mbox series

[FFmpeg-devel,1/3] avutil/mem: make max_alloc_size an atomic type

Message ID 20210522220904.7012-1-jamrial@gmail.com
State Accepted
Commit be96f4b616878c6245acd626e26cdd65a491b68d
Headers show
Series [FFmpeg-devel,1/3] avutil/mem: make max_alloc_size an atomic type | expand

Checks

Context Check Description
andriy/x86_make success Make finished
andriy/x86_make_fate success Make fate finished
andriy/PPC64_make success Make finished
andriy/PPC64_make_fate success Make fate finished

Commit Message

James Almer May 22, 2021, 10:09 p.m. UTC
This is in preparation for the following commit.

Signed-off-by: James Almer <jamrial@gmail.com>
---
 libavutil/mem.c | 17 +++++++++++------
 1 file changed, 11 insertions(+), 6 deletions(-)

Comments

Andreas Rheinhardt May 23, 2021, 2:14 p.m. UTC | #1
James Almer:
> This is in preparation for the following commit.
> 
> Signed-off-by: James Almer <jamrial@gmail.com>
> ---
>  libavutil/mem.c | 17 +++++++++++------
>  1 file changed, 11 insertions(+), 6 deletions(-)
> 
> diff --git a/libavutil/mem.c b/libavutil/mem.c
> index fa227f5e12..c12c24aa90 100644
> --- a/libavutil/mem.c
> +++ b/libavutil/mem.c
> @@ -31,6 +31,7 @@
>  #include <limits.h>
>  #include <stdint.h>
>  #include <stdlib.h>
> +#include <stdatomic.h>
>  #include <string.h>
>  #if HAVE_MALLOC_H
>  #include <malloc.h>
> @@ -68,17 +69,17 @@ void  free(void *ptr);
>   * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
>   * Note that this will cost performance. */
>  
> -static size_t max_alloc_size= INT_MAX;
> +static atomic_size_t max_alloc_size = ATOMIC_VAR_INIT(INT_MAX);
>  
>  void av_max_alloc(size_t max){
> -    max_alloc_size = max;
> +    atomic_store_explicit(&max_alloc_size, max, memory_order_relaxed);
>  }
>  
>  void *av_malloc(size_t size)
>  {
>      void *ptr = NULL;
>  
> -    if (size > max_alloc_size)
> +    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
>          return NULL;
>  
>  #if HAVE_POSIX_MEMALIGN
> @@ -134,7 +135,7 @@ void *av_malloc(size_t size)
>  void *av_realloc(void *ptr, size_t size)
>  {
>      void *ret;
> -    if (size > max_alloc_size)
> +    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
>          return NULL;
>  
>  #if HAVE_ALIGNED_MALLOC
> @@ -483,15 +484,19 @@ void av_memcpy_backptr(uint8_t *dst, int back, int cnt)
>  
>  void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size)
>  {
> +    size_t max_size;
> +
>      if (min_size <= *size)
>          return ptr;
>  
> -    if (min_size > max_alloc_size) {
> +    max_size = atomic_load_explicit(&max_alloc_size, memory_order_relaxed);
> +
> +    if (min_size > max_size) {
>          *size = 0;
>          return NULL;
>      }
>  
> -    min_size = FFMIN(max_alloc_size, FFMAX(min_size + min_size / 16 + 32, min_size));
> +    min_size = FFMIN(max_size, FFMAX(min_size + min_size / 16 + 32, min_size));
>  
>      ptr = av_realloc(ptr, min_size);
>      /* we could set this to the unmodified min_size but this is safer
> 
LGTM apart from the commit message.

- Andreas
James Almer May 23, 2021, 2:55 p.m. UTC | #2
On 5/23/2021 11:14 AM, Andreas Rheinhardt wrote:
> James Almer:
>> This is in preparation for the following commit.
>>
>> Signed-off-by: James Almer <jamrial@gmail.com>
>> ---
>>   libavutil/mem.c | 17 +++++++++++------
>>   1 file changed, 11 insertions(+), 6 deletions(-)
>>
>> diff --git a/libavutil/mem.c b/libavutil/mem.c
>> index fa227f5e12..c12c24aa90 100644
>> --- a/libavutil/mem.c
>> +++ b/libavutil/mem.c
>> @@ -31,6 +31,7 @@
>>   #include <limits.h>
>>   #include <stdint.h>
>>   #include <stdlib.h>
>> +#include <stdatomic.h>
>>   #include <string.h>
>>   #if HAVE_MALLOC_H
>>   #include <malloc.h>
>> @@ -68,17 +69,17 @@ void  free(void *ptr);
>>    * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
>>    * Note that this will cost performance. */
>>   
>> -static size_t max_alloc_size= INT_MAX;
>> +static atomic_size_t max_alloc_size = ATOMIC_VAR_INIT(INT_MAX);
>>   
>>   void av_max_alloc(size_t max){
>> -    max_alloc_size = max;
>> +    atomic_store_explicit(&max_alloc_size, max, memory_order_relaxed);
>>   }
>>   
>>   void *av_malloc(size_t size)
>>   {
>>       void *ptr = NULL;
>>   
>> -    if (size > max_alloc_size)
>> +    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
>>           return NULL;
>>   
>>   #if HAVE_POSIX_MEMALIGN
>> @@ -134,7 +135,7 @@ void *av_malloc(size_t size)
>>   void *av_realloc(void *ptr, size_t size)
>>   {
>>       void *ret;
>> -    if (size > max_alloc_size)
>> +    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
>>           return NULL;
>>   
>>   #if HAVE_ALIGNED_MALLOC
>> @@ -483,15 +484,19 @@ void av_memcpy_backptr(uint8_t *dst, int back, int cnt)
>>   
>>   void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size)
>>   {
>> +    size_t max_size;
>> +
>>       if (min_size <= *size)
>>           return ptr;
>>   
>> -    if (min_size > max_alloc_size) {
>> +    max_size = atomic_load_explicit(&max_alloc_size, memory_order_relaxed);
>> +
>> +    if (min_size > max_size) {
>>           *size = 0;
>>           return NULL;
>>       }
>>   
>> -    min_size = FFMIN(max_alloc_size, FFMAX(min_size + min_size / 16 + 32, min_size));
>> +    min_size = FFMIN(max_size, FFMAX(min_size + min_size / 16 + 32, min_size));
>>   
>>       ptr = av_realloc(ptr, min_size);
>>       /* we could set this to the unmodified min_size but this is safer
>>
> LGTM apart from the commit message.

Fixed commit message and applied, thanks.
Anton Khirnov May 31, 2021, 8:09 a.m. UTC | #3
Quoting James Almer (2021-05-23 00:09:02)
> This is in preparation for the following commit.
> 
> Signed-off-by: James Almer <jamrial@gmail.com>
> ---
>  libavutil/mem.c | 17 +++++++++++------
>  1 file changed, 11 insertions(+), 6 deletions(-)
> 
> diff --git a/libavutil/mem.c b/libavutil/mem.c
> index fa227f5e12..c12c24aa90 100644
> --- a/libavutil/mem.c
> +++ b/libavutil/mem.c
> @@ -31,6 +31,7 @@
>  #include <limits.h>
>  #include <stdint.h>
>  #include <stdlib.h>
> +#include <stdatomic.h>
>  #include <string.h>
>  #if HAVE_MALLOC_H
>  #include <malloc.h>
> @@ -68,17 +69,17 @@ void  free(void *ptr);
>   * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
>   * Note that this will cost performance. */
>  
> -static size_t max_alloc_size= INT_MAX;
> +static atomic_size_t max_alloc_size = ATOMIC_VAR_INIT(INT_MAX);
>  
>  void av_max_alloc(size_t max){
> -    max_alloc_size = max;
> +    atomic_store_explicit(&max_alloc_size, max, memory_order_relaxed);

Any specific reason for using a non-default memory order?
AFAIK it is recommended by the spec authors to use default (sequentially
consistent) operations unless there is a very good reason to do
something else.
James Almer May 31, 2021, 2:02 p.m. UTC | #4
On 5/31/2021 5:09 AM, Anton Khirnov wrote:
> Quoting James Almer (2021-05-23 00:09:02)
>> This is in preparation for the following commit.
>>
>> Signed-off-by: James Almer <jamrial@gmail.com>
>> ---
>>   libavutil/mem.c | 17 +++++++++++------
>>   1 file changed, 11 insertions(+), 6 deletions(-)
>>
>> diff --git a/libavutil/mem.c b/libavutil/mem.c
>> index fa227f5e12..c12c24aa90 100644
>> --- a/libavutil/mem.c
>> +++ b/libavutil/mem.c
>> @@ -31,6 +31,7 @@
>>   #include <limits.h>
>>   #include <stdint.h>
>>   #include <stdlib.h>
>> +#include <stdatomic.h>
>>   #include <string.h>
>>   #if HAVE_MALLOC_H
>>   #include <malloc.h>
>> @@ -68,17 +69,17 @@ void  free(void *ptr);
>>    * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
>>    * Note that this will cost performance. */
>>   
>> -static size_t max_alloc_size= INT_MAX;
>> +static atomic_size_t max_alloc_size = ATOMIC_VAR_INIT(INT_MAX);
>>   
>>   void av_max_alloc(size_t max){
>> -    max_alloc_size = max;
>> +    atomic_store_explicit(&max_alloc_size, max, memory_order_relaxed);
> 
> Any specific reason for using a non-default memory order?
> AFAIK it is recommended by the spec authors to use default (sequentially
> consistent) operations unless there is a very good reason to do
> something else.

I copied the existing behavior of av_force_cpu_flags() and 
av_get_cpu_flags(), which seems to be enough for a similar atomic 
variable as max_alloc_size.
diff mbox series

Patch

diff --git a/libavutil/mem.c b/libavutil/mem.c
index fa227f5e12..c12c24aa90 100644
--- a/libavutil/mem.c
+++ b/libavutil/mem.c
@@ -31,6 +31,7 @@ 
 #include <limits.h>
 #include <stdint.h>
 #include <stdlib.h>
+#include <stdatomic.h>
 #include <string.h>
 #if HAVE_MALLOC_H
 #include <malloc.h>
@@ -68,17 +69,17 @@  void  free(void *ptr);
  * dynamic libraries and remove -Wl,-Bsymbolic from the linker flags.
  * Note that this will cost performance. */
 
-static size_t max_alloc_size= INT_MAX;
+static atomic_size_t max_alloc_size = ATOMIC_VAR_INIT(INT_MAX);
 
 void av_max_alloc(size_t max){
-    max_alloc_size = max;
+    atomic_store_explicit(&max_alloc_size, max, memory_order_relaxed);
 }
 
 void *av_malloc(size_t size)
 {
     void *ptr = NULL;
 
-    if (size > max_alloc_size)
+    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
         return NULL;
 
 #if HAVE_POSIX_MEMALIGN
@@ -134,7 +135,7 @@  void *av_malloc(size_t size)
 void *av_realloc(void *ptr, size_t size)
 {
     void *ret;
-    if (size > max_alloc_size)
+    if (size > atomic_load_explicit(&max_alloc_size, memory_order_relaxed))
         return NULL;
 
 #if HAVE_ALIGNED_MALLOC
@@ -483,15 +484,19 @@  void av_memcpy_backptr(uint8_t *dst, int back, int cnt)
 
 void *av_fast_realloc(void *ptr, unsigned int *size, size_t min_size)
 {
+    size_t max_size;
+
     if (min_size <= *size)
         return ptr;
 
-    if (min_size > max_alloc_size) {
+    max_size = atomic_load_explicit(&max_alloc_size, memory_order_relaxed);
+
+    if (min_size > max_size) {
         *size = 0;
         return NULL;
     }
 
-    min_size = FFMIN(max_alloc_size, FFMAX(min_size + min_size / 16 + 32, min_size));
+    min_size = FFMIN(max_size, FFMAX(min_size + min_size / 16 + 32, min_size));
 
     ptr = av_realloc(ptr, min_size);
     /* we could set this to the unmodified min_size but this is safer