[FFmpeg-devel] [PATCH 1/3] common: add ff_parity()

[James Almer]

On 1/3/2016 5:33 PM, wm4 wrote:
> On Sun,  3 Jan 2016 20:21:00 +0100
> Clément Bœsch <u at pkh.me> wrote:
> 
>> ---
>>  configure               | 2 ++
>>  libavutil/x86/intmath.h | 9 +++++++++
>>  2 files changed, 11 insertions(+)
>>
>> diff --git a/configure b/configure
>> index 6710f85..610be92 100755
>> --- a/configure
>> +++ b/configure
>> @@ -1738,6 +1738,7 @@ BUILTIN_LIST="
>>      machine_rw_barrier
>>      MemoryBarrier
>>      mm_empty
>> +    parity
>>      rdtsc
>>      sarestart
>>      sync_val_compare_and_swap
>> @@ -5242,6 +5243,7 @@ check_builtin sarestart signal.h "SA_RESTART"
>>  check_builtin sync_val_compare_and_swap "" "int *ptr; int oldval, newval; __sync_val_compare_and_swap(ptr, oldval, newval)"
>>  check_builtin gmtime_r time.h "time_t *time; struct tm *tm; gmtime_r(time, tm)"
>>  check_builtin localtime_r time.h "time_t *time; struct tm *tm; localtime_r(time, tm)"
>> +check_builtin parity "" "__builtin_parity(123)"
>>  
>>  case "$custom_allocator" in
>>      jemalloc)
>> diff --git a/libavutil/x86/intmath.h b/libavutil/x86/intmath.h
>> index 611ef88..9c36bf2 100644
>> --- a/libavutil/x86/intmath.h
>> +++ b/libavutil/x86/intmath.h
>> @@ -75,6 +75,15 @@ static av_always_inline av_const int ff_ctzll_x86(long long v)
>>  
>>  #endif /* __POPCNT__ */
>>  
>> +static av_always_inline av_const int ff_parity(uint32_t v)
>> +{
>> +#if HAVE_PARITY
>> +    return __builtin_parity(v);
>> +#else
>> +    return av_popcount(v) & 1;
>> +#endif
> 
> Do compilers really generate better code for the former?

GCC does on x86 when the target cpu doesn't support the popcnt instruction,
otherwise the end result would be the same (popcnt + and).
av_popcount_c() is not optimal for this.

> 
>> +}
>> +
>>  #if defined(__BMI2__)
>>  
>>  #if AV_GCC_VERSION_AT_LEAST(5,1)
> 
> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
>