On 05/04/2012 03:50 PM, Måns Rullgård wrote:
The 16-bit FFT uses 32-bit intermediate values in many places, so it
does not process twice as much per instruction. It also has to do a few
shifts that the float version does not. It seems reasonable to me that
these implementations are roughly the
Orjan Friberg writes:
> Hi,
>
> Comparing the performance of fft-test and fft-fixed-test on a
> Cortex-A8 (with Neon support enabled) I only see a very small
> performance increase with the 16-bit fixed point version compared to
> the float version, regardless of the FFT size (64, 256, 4096). I
Hi,
Comparing the performance of fft-test and fft-fixed-test on a Cortex-A8
(with Neon support enabled) I only see a very small performance increase
with the 16-bit fixed point version compared to the float version,
regardless of the FFT size (64, 256, 4096). I didn't see any
performance num