Bug Fix

v1.2.1 contains following fixes:

• [dsp] Call appropriate NEON butterfly functions in NE10_fft_int32.neon.c
• [dsp] Prevent overflow by moving scaling in front of butterfly
• [dsp] Link stdc++ for static build
• [all] Switch to intrinsic on iOS
• [dsp] Fix NE10_INLINE_ASM_OPT for RFFT
• [math] Limit the range of input to avoid the failed tests in math module
• [dsp] Fix test_suite_fft_int32.c fails under armv7
  • For 4-point and 8-point FFTs, we fall back to VFP because they are too short for NEON optimizations
• [all] Main fuction in tests return 0 if all tests pass
• [all] Enable neon code on ARM32 in physics, dsp and image modules

It also contains a small improvement:

• [dsp] Improve data layout in RFFT float32

Add non-power-of-2 support for floating point CFFT

Radix-3 and radix-5 are supported for floating point CFFT

• C version supports almost all size of FFT
• NEON version supports radix {2, 3, 4, 5}, but size of FFT has to be multiple of 4
• c++ toolchain is required to build this library but all API remain pure C
• gcc 4.8 or higher is required to build module DSP, while gcc 4.9.3 or later is recommended