[FFmpeg-devel] [PATCH 5/5] aarch64/opusdsp: implement NEON accerelated postfilter and deemphasis

Carl Eugen Hoyos ceffmpeg at gmail.com
Sat Mar 23 18:20:18 EET 2019


2019-03-23 17:16 GMT+01:00, Lynne <dev at lynne.ee>:
> 23 Mar 2019, 15:04 by ceffmpeg at gmail.com:
>
>> 2019-03-23 15:23 GMT+01:00, Lynne <> dev at lynne.ee <mailto:dev at lynne.ee>>
>> >:
>>
>>> 16 Mar 2019, 16:34 by >> dev at lynne.ee <mailto:dev at lynne.ee>>> :
>>>
>>>> 153372 UNITS in postfilter_c,   65536 runs,      0 skips
>>>> 73164 UNITS in postfilter_neon,   65536 runs,      0 skips -> 2.1x
>>>> speedup
>>>>
>>>> 80591 UNITS in deemphasis_c,  131072 runs,      0 skips
>>>> 43969 UNITS in deemphasis_neon,  131072 runs,      0 skips -> 1.83x
>>>> speedup
>>>>
>>>> Total decoder speedup: ~15% on a Raspberry Pi 3 (from 28.1x to 33.5x
>>>> realtime)
>>>>
>>>> Deemphasis SIMD based on the following unrolling:
>>>> const float c1 = CELT_EMPH_COEFF, c2 = c1*c1, c3 = c2*c1, c4 = c3*c1;
>>>> float state = coeff;
>>>>
>>>> for (int i = 0; i < len; i += 4) {
>>>>  y[0] = x[0] + c1*state;
>>>>  y[1] = x[1] + c2*state + c1*x[0];
>>>>  y[2] = x[2] + c3*state + c1*x[1] + c2*x[0];
>>>>  y[3] = x[3] + c4*state + c1*x[2] + c2*x[1] + c3*x[0];
>>>>
>>>>  state = y[3];
>>>>  y += 4;
>>>>  x += 4;
>>>> }
>>>>
>>>> Unlike the x86 version, duplication is used instead of pslldq so
>>>> the structure and tables are different.
>>>> Same approach tested on x86 (3x pslldq -> vbroadcastss + shufps +
>>>> pslldq)
>>>> had the same performance, so 3x pslldq was kept as vbroadcastss has a
>>>> higher latency.
>>>>
>>>
>>> Could someone review the patches?
>>>
>>
>> Which toolchains did you test?
>> (For compilation, not performance.)
>>
>
> gcc 8.2.1 on both aarch64 and x86-64

Please also test Android and tell us if you
can test ios compilation.
(Assuming you cannot test arm64 for Windows.)

Carl Eugen


More information about the ffmpeg-devel mailing list