Matrix multiplication functions that were recently marked as 'constexpr'
in commit '1cc8e80e3ba140239196d9a4597a2ea8139a4aa5' can throw warnings
about an "uninitialized variable 'Result' in constexpr function".
Change-Id: I95396da9ac8a6e0dd1b6ae4e782f75446cfa70a3
Android NDK starting from r18 has libcxx, and it's the only STL available. (reference: https://developer.android.com/ndk/guides/cpp-support#cs)
r18 is ~6 years old at this point.
This check dates to ~8 years ago (predating r18) and was important at that time.
Right now, it can be clearly stated that given the C++11 requirement stated in README for GLM, as well as (unless you are building with very outdated toolchain) - all modern Android is built with NDK toolchain that is newer than r18 - this check can be removed, and all the functionality can by default delegate to general detection for STL compatibility.
- Add simd aligned_vec3 (and sse aligned_dvec3 - 2 x xmm)
- Fast packed_vec3 <=> aligned_vec3 and packed_vec4 <=> aligned_vec4 conversion
- Fast aligned_vec3 <=> aligned_vec4 conversion
- Optimized aligned_mat x aligned_mat and aligned_mat x aligned_vec
- Inverse aligned_mat3 simd version (actually slower than ssid on my computer even it has 30% less instruction ?)
- Add simd aligned_vec3 (and sse aligned_dvec3 - 2 x xmm)
- Fast packed_vec3 <=> aligned_vec3 and packed_vec4 <=> aligned_vec4 conversion
- Fast aligned_vec3 <=> aligned_vec4 conversion
- Optimized aligned_mat x aligned_mat and aligned_mat x aligned_vec
- Inverse aligned_mat3 simd version (actually slower than ssid on my computer even it has 30% less instruction ?)
Depending on the signedness of length_t type we may either get
warnings comparison of unsigned expression in ‘>= 0’ being always true
or do insufficient checking by not checking if index is not negative.
Hide the index checking behind a macro that check the index properly.
