Hi,

On Thu, Jun 2, 2011 at 3:15 PM, John Stebbins <[email protected]> wrote:
> On 04/26/2011 09:28 AM, Vitor Sessak wrote:
>>
>> Module: libav
>> Branch: master
>> Commit: 13dfce3d44f99a2d7df71aba8ae003d58db726f7
>>
>> Author:    Vitor Sessak<[email protected]>
>> Committer: Reinhard Tartler<[email protected]>
>> Date:      Sat Apr 23 19:24:31 2011 +0200
>>
>> Increase alignment of av_malloc() as needed by AVX ASM.
>>
>> Signed-off-by: Reinhard Tartler<[email protected]>
>>
>> ---
>>
>>  libavutil/mem.c |   16 +++++++---------
>>  1 files changed, 7 insertions(+), 9 deletions(-)
>>
>> diff --git a/libavutil/mem.c b/libavutil/mem.c
>> index 2aef9b0..27bb30b 100644
>> --- a/libavutil/mem.c
>> +++ b/libavutil/mem.c
>> @@ -69,21 +69,21 @@ void *av_malloc(size_t size)
>>  #endif
>>
>>      /* let's disallow possible ambiguous cases */
>> -    if(size>  (INT_MAX-16) )
>> +    if(size>  (INT_MAX-32) )
>>          return NULL;
>>
>>  #if CONFIG_MEMALIGN_HACK
>> -    ptr = malloc(size+16);
>> +    ptr = malloc(size+32);
>>      if(!ptr)
>>          return ptr;
>> -    diff= ((-(long)ptr - 1)&15) + 1;
>> +    diff= ((-(long)ptr - 1)&31) + 1;
>>      ptr = (char*)ptr + diff;
>>      ((char*)ptr)[-1]= diff;
>>  #elif HAVE_POSIX_MEMALIGN
>> -    if (posix_memalign(&ptr,16,size))
>> +    if (posix_memalign(&ptr,32,size))
>>          ptr = NULL;
>>  #elif HAVE_MEMALIGN
>> -    ptr = memalign(16,size);
>> +    ptr = memalign(32,size);
>>      /* Why 64?
>>         Indeed, we should align it:
>>           on 4 for 386
>> @@ -93,10 +93,8 @@ void *av_malloc(size_t size)
>>         Because L1 and L2 caches are aligned on those values.
>>         But I don't want to code such logic here!
>>       */
>> -     /* Why 16?
>> -        Because some CPUs need alignment, for example SSE2 on P4,&  most
>> RISC CPUs
>> -        it will just trigger an exception and the unaligned load will be
>> done in the
>> -        exception handler or it will just segfault (SSE2 on P4).
>> +     /* Why 32?
>> +        For AVX ASM. SSE / NEON needs only 16.
>>          Why not larger? Because I did not see a difference in benchmarks
>> ...
>>       */
>>       /* benchmarks with P3
>>
>>
> This commit breaks OSX on SDK 10.6 which has a broken posix_memalign
> implementation. /Ronald Bultje /has mentioned this OSX bug in other threads
> on this list. When alignment >=32 && size == 0, posix_memalign returns an
> invalid pointer along with a success code. So subsequent free corrupts the
> heap. Something like the attached patch fixes the issue.
>

I've tried to fix all occurrences of this in the code. I don't want to
put a workaround in av_malloc(), I think it's better not to call the
function when size==0 anyway.

(Not to mention OSX is broken and Apple isn't responding to my bug report.)

Ronald
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel

Reply via email to