mufft benchmark problem

[setoutsoft]

Recently, I had tried to incorporate mufft into our project. Unfortunately, we found, if we disable sse or avx, mufft is much slower than other fft implements such as kiss fft, small fft etc.
For our test samples, benchmark result will like below:

release version (ms):
kiss fft: 1461
small fft: 1447
mufft with disabled sse :1735
mufft with sse3: 1234

both kill fft and small fft don't have been optimized by sse.
Do you do the same benchmark?

[setoutsoft]

The problem is I do the benchmark using debug mode. Switching to release mode, with neon support, mufft is fast than kiss fft for about 10%.