AlanCui

@christophe0606 Sorry for my mistake, it's ok to include arm_neon.h on M4F, even there is only a subset of NEON implemented on M4F, but the vcvt_s32_f32 will call a SIMD...

@christophe0606 I figure out it, ACLE have no single precision version (FPv4-**SP**) for vcvt_s32_f32, it's either d-register or q-register in ACLE. So I do believe someone forget to add it...

@christophe0606 Hi, Is that ok to add a inline assembly specilized for M4? If so, i will open a pull request.

> @christophe0606, @AlanCui4080, where does this new instrinsic belong to? Is this one the correct location? https://github.com/ARM-software/CMSIS_6/blob/8c4dc58928b3347f6aa98b6fb2bf6770f32a72b7/CMSIS/Core/Include/cmsis_gcc.h#L877-L992 @JonatanAntoni No, that is not a DSP instruction, it's belong to FPU. I...

> So this assertion is totally untrue: > > > since there is no absolute value statement, square root statement, or q31 to float statement in C language, compliers will...

@christophe0606 > As consequence, and unfortunately, it is dependent on what compilers can do. But I see many manually vectorizing in code. Anyway, as it can be auto optimized by...

Closed, see https://github.com/ARM-software/CMSIS-DSP/issues/212.