@@ -401,7 +401,6 @@
reciRoot = MUL_F32_SIMD(MUL_F32_SIMD(LOAD_SCALAR_F32_SIMD(0.5f), reciRoot), SUB_F32_SIMD(LOAD_SCALAR_F32_SIMD(3.0f), mul));
return F32x4(reciRoot);
#elif USE_NEON
- // TODO: Test on ARM
// Approximate
SIMD_F32x4 reciRoot = vrsqrteq_f32(this->v);
// Refine
@@ -427,7 +426,6 @@
root = _mm_mul_ps(_mm_add_ps(root, _mm_div_ps(this->v, root)), half);
return F32x4(root);
return F32x4(MUL_F32_SIMD(this->v, this->reciprocalSquareRoot().v));
#else
assert(false);