[FFmpeg-devel] [PATCH 4/5] avcodec/fft_template: Only check for FF_FFT_PERM_AVX on ARCH_X86

Lynne dev at lynne.ee
Thu Jan 7 19:50:28 EET 2021


Jan 7, 2021, 17:38 by andreas.rheinhardt at gmail.com:

> Lynne:
>
>> Jan 7, 2021, 00:13 by andreas.rheinhardt at gmail.com:
>>
>>> Also do it for FFT_FLOAT only, as this is the only combination for which
>>> it can be set.
>>>
>>> Signed-off-by: Andreas Rheinhardt <andreas.rheinhardt at gmail.com>
>>> ---
>>>  libavcodec/fft_template.c | 2 +-
>>>  1 file changed, 1 insertion(+), 1 deletion(-)
>>>
>>> diff --git a/libavcodec/fft_template.c b/libavcodec/fft_template.c
>>> index 9d125de073..ddde63714e 100644
>>> --- a/libavcodec/fft_template.c
>>> +++ b/libavcodec/fft_template.c
>>> @@ -251,7 +251,7 @@ av_cold int ff_fft_init(FFTContext *s, int nbits, int inverse)
>>>  #endif /* FFT_FIXED_32 */
>>>  
>>>  
>>> -    if (s->fft_permutation == FF_FFT_PERM_AVX) {
>>> +    if (ARCH_X86 && FFT_FLOAT && s->fft_permutation == FF_FFT_PERM_AVX) {
>>>  fft_perm_avx(s);
>>>  } else {
>>>  #define PROCESS_FFT_PERM_SWAP_LSBS(num) do {\
>>>
>>
>> LGTM. Maybe mark fft_perm_avx as inline too if you can be bothered to amend the patch.
>>
>
> I don't see a reason to interfere in the compiler's inlining decision
> here. It is a static function only called once, so it will be inlined
> anyway.
>

Fair enough.


More information about the ffmpeg-devel mailing list