[FFmpeg-devel] [PATCH] aacenc_utils: unroll loops to allow compiler to use SIMD.

[Ganesh Ajjanagadde]

On Tue, Mar 8, 2016 at 2:30 PM, Reimar Döffinger
<Reimar.Doeffinger at gmx.de> wrote:
> On Mon, Mar 07, 2016 at 10:50:53PM -0500, Ganesh Ajjanagadde wrote:
>> On Mon, Mar 7, 2016 at 2:54 AM, Reimar Döffinger
>> <Reimar.Doeffinger at gmx.de> wrote:
>> >> Can you be more specific, and are you sure about this?
>> >
>> > Just run your favourite performance analysis tool and you'll see.
>> > As it is non-inlined libc code I'm fairly sure the numbers are accurate enough.
>>
>> I am still puzzled by the remarks; and hence asked for specific
>> examples. In aac code, cosf is only called for table generation, same
>> with cos, so still don't see why cos is relevant.
>
> You seem to have missed ff_lpc_calc_ref_coefs_f.
> At least it's that function that is to blame for
> the cos calls I saw in the profiler.

Ah, now I understand - ff_lpc_calc_refs_coefs_f' is defined in
lavc/lpc, I was only looking internally at lavc/aac*.
Thanks for the heads up.

Seems like cos() symmetry can be exploited at the very least. Does not
seem like len is a power of 2, so can't directly use cos tables.

> _______________________________________________
> ffmpeg-devel mailing list
> ffmpeg-devel at ffmpeg.org
> http://ffmpeg.org/mailman/listinfo/ffmpeg-devel