common/math fails to compile on linux on ARM

[ap2468]

The following MWE fails to compile due to a couple issues.

// file test.cpp
#include <iostream>
#include <math/vec3.h>
#include <math/vec3fa.h>

int main() {
    embree::Vec3fa vec(1.0,2.0,3.0);
    std::cout << vec << std::endl;
    return 0;
}

compiled with g++ -I./embree/common -I./embree/include -O3 -DNDEBUG -std=gnu++17 ./test.cpp -o test results in a couple errors:

1. Missing definitions:

In file included from ./embree/common/math/vec3.h:262,
                 from ./test.cpp:4:
./embree/common/math/vec3fa.h: At global scope:
./embree/common/math/vec3fa.h:52:44: error: ISO C++ forbids declaration of ‘type name’ with no type [-fpermissive]
   52 |     __forceinline explicit operator const  Vec2fa() const { return Vec2fa(m128); }
      |                                            ^~~~~~
./embree/common/math/vec3fa.h:52:28: error: ‘operator const int’ does not name a type
   52 |     __forceinline explicit operator const  Vec2fa() const { return Vec2fa(m128); }
      |                            ^~~~~~~~
./embree/common/math/vec3fa.h:474:44: error: ISO C++ forbids declaration of ‘type name’ with no type [-fpermissive]
  474 |     __forceinline explicit operator const  Vec2fa() const { return Vec2fa(m128); }
      |                                            ^~~~~~
./embree/common/math/vec3fa.h:474:28: error: ‘operator const int’ does not name a type
  474 |     __forceinline explicit operator const  Vec2fa() const { return Vec2fa(m128); }
      |                            ^~~~~~~~

2. some conversion errors, e.g.:

In file included from ./embree/common/math/../sys/../simd/arm/emulation.h:27,
                 from ./embree/common/math/../sys/intrinsics.h:13,
                 from ./embree/common/math/emath.h:7,
                 from ./embree/common/math/vec3.h:6,
                 from ./test.cpp:4:
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_setzero_si256()’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:161:36: note: use ‘-flax-vector-conversions’ to permit conversions between vectors with differing element types or numbers of subparts
  161 |     res.lo = res.hi = vdupq_n_s32(0);
      |                                    ^
./embree/common/math/../sys/../simd/arm/avx2neon.h:161:34: error: cannot convert ‘int32x4_t’ to ‘__m128i’ {aka ‘int64x2_t’} in assignment
  161 |     res.lo = res.hi = vdupq_n_s32(0);
      |                       ~~~~~~~~~~~^~~
      |                                  |
      |                                  int32x4_t
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_castsi128_si256(__m128i)’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:206:25: error: cannot convert ‘int32x4_t’ to ‘__m128i’ {aka ‘int64x2_t’} in assignment
  206 |     res.hi = vdupq_n_s32(0);
      |              ~~~~~~~~~~~^~~
      |                         |
      |                         int32x4_t
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_set1_epi32(int)’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:249:34: error: cannot convert ‘int32x4_t’ to ‘__m128i’ {aka ‘int64x2_t’} in assignment
  249 |     res.lo = res.hi = vdupq_n_s32(a);
      |                       ~~~~~~~~~~~^~~
      |                                  |
      |                                  int32x4_t
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_set1_epi8(int)’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:256:33: error: cannot convert ‘int8x16_t’ to ‘__m128i’ {aka ‘int64x2_t’} in assignment
  256 |     res.lo = res.hi = vdupq_n_s8(a);
      |                       ~~~~~~~~~~^~~
      |                                 |
      |                                 int8x16_t
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_set1_epi16(int)’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:263:34: error: cannot convert ‘int16x8_t’ to ‘__m128i’ {aka ‘int64x2_t’} in assignment
  263 |     res.lo = res.hi = vdupq_n_s16(a);
      |                       ~~~~~~~~~~~^~~
      |                                  |
      |                                  int16x8_t
./embree/common/math/../sys/../simd/arm/avx2neon.h: In function ‘__m256i _mm256_abs_epi32(__m256i)’:
./embree/common/math/../sys/../simd/arm/avx2neon.h:436:25: error: cannot convert ‘__m128i’ {aka ‘int64x2_t’} to ‘int32x4_t’
  436 |    res.lo = vabsq_s32(a.lo);
      |                       ~~^~
      |                         |
      |                         __m128i {aka int64x2_t}

The first error can be fixed by including

#include <math/vec2.h>
#include <math/vec2fa.h>

Should these be included in vec3.h?

The second can be fixed by adding -flax-vector-conversions to the compile flags, but maybe the fixes that were used to address #410 in the main library can also be applied to the code in common?