feat: misc additions (#221)

nfrechette · Sep 7, 2024 · 4466148 · 4466148
2 parents 14416d4 + 38fca19
commit 4466148
Show file tree

Hide file tree

Showing 12 changed files with 514 additions and 0 deletions.
diff --git a/includes/rtm/impl/mask_common.h b/includes/rtm/impl/mask_common.h
@@ -259,6 +259,22 @@ namespace rtm
 		return rtm_impl::mask4_uint64_set{ x, y, z, w };
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Creates a mask4 with all 4 components set to true.
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE constexpr rtm_impl::mask4_bool_set RTM_SIMD_CALL mask_true() RTM_NO_EXCEPT
+	{
+		return rtm_impl::mask4_bool_set{ true, true, true, true };
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Creates a mask4 with all 4 components set to false.
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE constexpr rtm_impl::mask4_bool_set RTM_SIMD_CALL mask_false() RTM_NO_EXCEPT
+	{
+		return rtm_impl::mask4_bool_set{ false, false, false, false };
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/mask4d.h b/includes/rtm/mask4d.h
@@ -400,6 +400,34 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component logical NOT of the input: ~input
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE mask4d RTM_SIMD_CALL mask_not(mask4d_arg0 input) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		const __m128i true_mask = _mm_set_epi64x(0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL);
+		__m128d xy = _mm_andnot_pd(input.xy, _mm_castsi128_pd(true_mask));
+		__m128d zw = _mm_andnot_pd(input.zw, _mm_castsi128_pd(true_mask));
+		return mask4d{ xy, zw };
+#else
+		const uint64_t* input_ = rtm_impl::bit_cast<const uint64_t*>(&input);
+
+		union
+		{
+			mask4d vector;
+			uint64_t scalar[4];
+		} result;
+
+		result.scalar[0] = ~input_[0];
+		result.scalar[1] = ~input_[1];
+		result.scalar[2] = ~input_[2];
+		result.scalar[3] = ~input_[3];
+
+		return result.vector;
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/mask4f.h b/includes/rtm/mask4f.h
@@ -359,6 +359,34 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component logical NOT of the input: ~input
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE mask4f RTM_SIMD_CALL mask_not(mask4f_arg0 input) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		const __m128i true_mask = _mm_set_epi32(0xFFFFFFFFULL, 0xFFFFFFFFULL, 0xFFFFFFFFULL, 0xFFFFFFFFULL);
+		return _mm_andnot_ps(input, _mm_castsi128_ps(true_mask));
+#elif defined(RTM_NEON_INTRINSICS)
+		return vmvnq_u32(input);
+#else
+		const uint32_t* input_ = rtm_impl::bit_cast<const uint32_t*>(&input);
+
+		union
+		{
+			mask4f vector;
+			uint32_t scalar[4];
+		} result;
+
+		result.scalar[0] = ~input_[0];
+		result.scalar[1] = ~input_[1];
+		result.scalar[2] = ~input_[2];
+		result.scalar[3] = ~input_[3];
+
+		return result.vector;
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/mask4i.h b/includes/rtm/mask4i.h
@@ -327,6 +327,21 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component logical NOT of the input: ~input
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE mask4i RTM_SIMD_CALL mask_not(mask4i_arg0 input) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		const __m128i true_mask = _mm_set_epi32(0xFFFFFFFFULL, 0xFFFFFFFFULL, 0xFFFFFFFFULL, 0xFFFFFFFFULL);
+		return _mm_andnot_si128(input, true_mask);
+#elif defined(RTM_NEON_INTRINSICS)
+		return RTM_IMPL_MASK4i_SET(vmvnq_u32(RTM_IMPL_MASK4i_GET(input)));
+#else
+		return mask4i{ ~input.x, ~input.y, ~input.z, ~input.w };
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/mask4q.h b/includes/rtm/mask4q.h
@@ -364,6 +364,34 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component logical NOT of the input: ~input
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE mask4q RTM_SIMD_CALL mask_not(mask4q_arg0 input) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		const __m128i true_mask = _mm_set_epi64x(0xFFFFFFFFFFFFFFFFULL, 0xFFFFFFFFFFFFFFFFULL);
+		__m128i xy = _mm_andnot_si128(input.xy, true_mask);
+		__m128i zw = _mm_andnot_si128(input.zw, true_mask);
+		return mask4q{ xy, zw };
+#else
+		const uint64_t* input_ = rtm_impl::bit_cast<const uint64_t*>(&input);
+
+		union
+		{
+			mask4q vector;
+			uint64_t scalar[4];
+		} result;
+
+		result.scalar[0] = ~input_[0];
+		result.scalar[1] = ~input_[1];
+		result.scalar[2] = ~input_[2];
+		result.scalar[3] = ~input_[3];
+
+		return result.vector;
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/quatd.h b/includes/rtm/quatd.h
@@ -1051,6 +1051,22 @@ namespace rtm
 		return positive_w_angle <= threshold_angle;
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component selection depending on the mask: mask != 0 ? if_true : if_false
+	// Note that if the mask lanes are not all identical, the resulting quaternion
+	// may not be normalized.
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE quatd RTM_SIMD_CALL quat_select(mask4d_arg0 mask, quatd_arg1 if_true, quatd_arg2 if_false) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy = RTM_VECTOR2D_SELECT(mask.xy, if_true.xy, if_false.xy);
+		__m128d zw = RTM_VECTOR2D_SELECT(mask.zw, if_true.zw, if_false.zw);
+		return quatd{ xy, zw };
+#else
+		return quatd{ rtm_impl::select(mask.x, if_true.x, if_false.x), rtm_impl::select(mask.y, if_true.y, if_false.y), rtm_impl::select(mask.z, if_true.z, if_false.z), rtm_impl::select(mask.w, if_true.w, if_false.w) };
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/quatf.h b/includes/rtm/quatf.h
@@ -1548,6 +1548,20 @@ namespace rtm
 		return positive_w_angle <= threshold_angle;
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Per component selection depending on the mask: mask != 0 ? if_true : if_false
+	// Note that if the mask lanes are not all identical, the resulting quaternion
+	// may not be normalized.
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE quatf RTM_SIMD_CALL quat_select(mask4f_arg0 mask, quatf_arg1 if_true, quatf_arg2 if_false) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS) || defined(RTM_NEON_INTRINSICS)
+		return RTM_VECTOR4F_SELECT(mask, if_true, if_false);
+#else
+		return quatf{ rtm_impl::select(mask.x, if_true.x, if_false.x), rtm_impl::select(mask.y, if_true.y, if_false.y), rtm_impl::select(mask.z, if_true.z, if_false.z), rtm_impl::select(mask.w, if_true.w, if_false.w) };
+#endif
+	}
+
 	RTM_IMPL_VERSION_NAMESPACE_END
 }
 

diff --git a/includes/rtm/vector4d.h b/includes/rtm/vector4d.h
@@ -2088,6 +2088,20 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Returns per component ~0 if not equal, otherwise 0: lhs != rhs ? ~0 : 0
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE mask4d RTM_SIMD_CALL vector_not_equal(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_lt_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		__m128d zw_lt_pd = _mm_cmpneq_pd(lhs.zw, rhs.zw);
+		return mask4d{ xy_lt_pd, zw_lt_pd };
+#else
+		return mask4d{ rtm_impl::get_mask_value(lhs.x != rhs.x), rtm_impl::get_mask_value(lhs.y != rhs.y), rtm_impl::get_mask_value(lhs.z != rhs.z), rtm_impl::get_mask_value(lhs.w != rhs.w) };
+#endif
+	}
+
 	//////////////////////////////////////////////////////////////////////////
 	// Returns per component ~0 if less than, otherwise 0: lhs < rhs ? ~0 : 0
 	//////////////////////////////////////////////////////////////////////////
@@ -2554,6 +2568,88 @@ namespace rtm
 #endif
 	}
 
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if all [xyzw] components are not equal, otherwise false: all(lhs.xyzw != rhs.xyzw)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_all_not_equal(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		__m128d zw_eq_pd = _mm_cmpneq_pd(lhs.zw, rhs.zw);
+		return (_mm_movemask_pd(xy_eq_pd) & _mm_movemask_pd(zw_eq_pd)) == 3;
+#else
+		return lhs.x != rhs.x && lhs.y != rhs.y && lhs.z != rhs.z && lhs.w != rhs.w;
+#endif
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if all [xy] components are not equal, otherwise false: all(lhs.xy != rhs.xy)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_all_not_equal2(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		return _mm_movemask_pd(xy_eq_pd) == 3;
+#else
+		return lhs.x != rhs.x && lhs.y != rhs.y;
+#endif
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if all [xyz] components are not equal, otherwise false: all(lhs.xyz != rhs.xyz)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_all_not_equal3(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		__m128d zw_eq_pd = _mm_cmpneq_pd(lhs.zw, rhs.zw);
+		return _mm_movemask_pd(xy_eq_pd) == 3 && (_mm_movemask_pd(zw_eq_pd) & 1) != 0;
+#else
+		return lhs.x != rhs.x && lhs.y != rhs.y && lhs.z != rhs.z;
+#endif
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if any [xyzw] components are not equal, otherwise false: any(lhs.xyzw != rhs.xyzw)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_any_not_equal(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		__m128d zw_eq_pd = _mm_cmpneq_pd(lhs.zw, rhs.zw);
+		return (_mm_movemask_pd(xy_eq_pd) | _mm_movemask_pd(zw_eq_pd)) != 0;
+#else
+		return lhs.x != rhs.x || lhs.y != rhs.y || lhs.z != rhs.z || lhs.w != rhs.w;
+#endif
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if any [xy] components are not equal, otherwise false: any(lhs.xy != rhs.xy)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_any_not_equal2(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		return _mm_movemask_pd(xy_eq_pd) != 0;
+#else
+		return lhs.x != rhs.x || lhs.y != rhs.y;
+#endif
+	}
+
+	//////////////////////////////////////////////////////////////////////////
+	// Returns true if any [xyz] components are not equal, otherwise false: any(lhs.xyz != rhs.xyz)
+	//////////////////////////////////////////////////////////////////////////
+	RTM_DISABLE_SECURITY_COOKIE_CHECK RTM_FORCE_INLINE bool RTM_SIMD_CALL vector_any_not_equal3(vector4d_arg0 lhs, vector4d_arg1 rhs) RTM_NO_EXCEPT
+	{
+#if defined(RTM_SSE2_INTRINSICS)
+		__m128d xy_eq_pd = _mm_cmpneq_pd(lhs.xy, rhs.xy);
+		__m128d zw_eq_pd = _mm_cmpneq_pd(lhs.zw, rhs.zw);
+		return _mm_movemask_pd(xy_eq_pd) != 0 || (_mm_movemask_pd(zw_eq_pd) & 1) != 0;
+#else
+		return lhs.x != rhs.x || lhs.y != rhs.y || lhs.z != rhs.z;
+#endif
+	}
+
 	//////////////////////////////////////////////////////////////////////////
 	// Returns true if all 4 components are near equal, otherwise false: all(abs(lhs - rhs).xyzw <= threshold)
 	//////////////////////////////////////////////////////////////////////////