Fix conversion from simd8f to simd8i for MSVC ARM on Windows #3039
Conversation
| #if defined(DLIB_HAVE_NEON) | ||
| temp.low = simd4i(vcvtq_s32_f32(_low)); | ||
| temp.high = simd4i(vcvtq_s32_f32(_high)); | ||
| #else |
There was a problem hiding this comment.
What happens exactly? It seems like this shouldn't be needed, since simd4f (which is what _low and _high are) has this already: inline operator int32x4_t() const { return vcvtq_s32_f32(x); }. And then simd4i has has this constructor as well inline simd4i(const int32x4_t& val):x(val) {}. So this change shouldn't be needed. But I guess visual studio is throwing some error? What's the error exactly?
There was a problem hiding this comment.
You are completely right. For whatever reason the operator inline operator int32x4_t() was missing on my side. Actually it compiled just fine, only it resulted in a crash at runtime in fhog.h. I will close this pull request.
|
Closed because it is a problem on my machine. |
This is needed because the conversion between float32x4_t to int32x4_t needs to be done explicit.