FastLED 3.9.15
Loading...
Searching...
No Matches

◆ FASTLED_DEFINE_POD_COMPARISON_OPERATOR

#define FASTLED_DEFINE_POD_COMPARISON_OPERATOR ( CLASS,
OP )
Value:
template <typename T, typename U> \
typename enable_if< \
is_same<U, CLASS>::value && is_pod<T>::value, bool>::type \
operator OP(const T &pod, const CLASS &obj) FL_NOEXCEPT { \
return pod OP obj; \
} \
template <typename T> \
typename enable_if<is_pod<T>::value, bool>::type operator OP( \
const CLASS &obj, const T &pod) FL_NOEXCEPT { \
return obj OP pod; \
}
#define FL_NOEXCEPT

Definition at line 1020 of file s16x16x4.h.

1020#pragma once
1021
1024
1025#include "fl/math/simd.h"
1027#include "fl/math/sin32.h"
1029#include "fl/stl/align.h"
1030
1031namespace fl {
1032
1033// Forward declaration for cross-type operations
1034struct s0x32x4;
1035
1038struct s16x16x4 {
1039 simd::simd_u32x4 raw; // 4× i32 values in Q16 format
1040
1041 // ---- Construction ------------------------------------------------------
1042
1044 s16x16x4 result;
1045 result.raw = r;
1046 return result;
1047 }
1048
1049 // Load 4 s16x16 values from memory (unaligned access supported)
1050 static FASTLED_FORCE_INLINE s16x16x4 load(const s16x16* ptr) {
1051 return from_raw(simd::platforms::load_u32_4(reinterpret_cast<const u32*>(ptr))); // ok reinterpret cast
1052 }
1053
1054 // Store 4 s16x16 values to memory (unaligned access supported)
1055 FASTLED_FORCE_INLINE void store(s16x16* ptr) const {
1056 simd::platforms::store_u32_4(reinterpret_cast<u32*>(ptr), raw); // ok reinterpret cast
1057 }
1058
1059 // Broadcast single s16x16 value to all 4 lanes
1060 static FASTLED_FORCE_INLINE s16x16x4 set1(s16x16 value) {
1061 return from_raw(simd::platforms::set1_u32_4(static_cast<u32>(value.raw())));
1062 }
1063
1064 // ---- SIMD arithmetic (s16x16x4 OP s16x16x4 → s16x16x4) -----------------
1065
1066 FASTLED_FORCE_INLINE s16x16x4 operator+(s16x16x4 b) const {
1067 return from_raw(simd::add_i32_4(raw, b.raw));
1068 }
1069
1070 FASTLED_FORCE_INLINE s16x16x4 operator-(s16x16x4 b) const {
1071 return from_raw(simd::sub_i32_4(raw, b.raw));
1072 }
1073
1074 FASTLED_FORCE_INLINE s16x16x4 operator*(s16x16x4 b) const {
1075 // Q16 × Q16 = Q32 → shift right 16 → Q16
1076 return from_raw(simd::mulhi_i32_4(raw, b.raw));
1077 }
1078
1079 FASTLED_FORCE_INLINE s16x16x4 operator-() const {
1080 // Unary negation: -x = 0 - x
1081 auto zero = simd::set1_u32_4(0);
1082 return from_raw(simd::sub_i32_4(zero, raw));
1083 }
1084
1085 FASTLED_FORCE_INLINE s16x16x4 operator>>(int shift) const {
1086 return from_raw(simd::sra_i32_4(raw, shift));
1087 }
1088
1089 FASTLED_FORCE_INLINE s16x16x4 operator<<(int shift) const {
1090 return from_raw(simd::sll_u32_4(raw, shift));
1091 }
1092
1093 // Cross-type multiply: s16x16x4 × s0x32x4 → s16x16x4 (commutative)
1094 // Implemented after s0x32x4 is defined
1095 FASTLED_FORCE_INLINE s16x16x4 operator*(s0x32x4 b) const;
1096
1097 // ---- Math functions -------------------------------------------------------
1098
1100 FASTLED_FORCE_INLINE s16x16x4 abs() const {
1101 // mask = -1 if negative (sign extended), 0 if positive
1102 auto mask = simd::sra_i32_4(raw, 31);
1103 // flip bits if negative, then add 1 (two's complement)
1104 auto flipped = simd::xor_u32_4(raw, mask);
1105 return from_raw(simd::sub_i32_4(flipped, mask));
1106 }
1107
1109 FASTLED_FORCE_INLINE s16x16x4 min(s16x16x4 b) const {
1110 return from_raw(simd::min_i32_4(raw, b.raw));
1111 }
1112
1114 FASTLED_FORCE_INLINE s16x16x4 max(s16x16x4 b) const {
1115 return from_raw(simd::max_i32_4(raw, b.raw));
1116 }
1117
1119 FASTLED_FORCE_INLINE s16x16x4 clamp(s16x16x4 lo, s16x16x4 hi) const {
1120 return max(lo).min(hi);
1121 }
1122
1125 FASTLED_FORCE_INLINE s16x16x4 lerp(s16x16x4 b, s16x16 t) const {
1126 auto t_vec = s16x16x4::set1(t);
1127 auto diff = b - (*this);
1128 return (*this) + (diff * t_vec);
1129 }
1130
1133 FASTLED_FORCE_INLINE void sincos(s16x16x4& out_sin, s16x16x4& out_cos) const {
1134 // Convert radians to 24-bit angle units (same as scalar s16x16)
1135 // RAD_TO_24 = 2^24 / (2π) in Q16
1136 static constexpr i32 RAD_TO_24 = 2670177; // from s16x16.h
1137
1138 // Convert 4 angles: mulhi_i32_4 does (i64*i64) >> 16
1139 auto angles_u32 = simd::mulhi_su32_4(raw, simd::set1_u32_4(static_cast<u32>(RAD_TO_24)));
1140
1141 // Call vectorized sincos
1142 auto sc = sincos32_simd(angles_u32);
1143
1144 // Shift results right by 15 to convert from raw sin32 output to Q16.16
1145 out_sin = from_raw(simd::sra_i32_4(sc.sin_vals, 15));
1146 out_cos = from_raw(simd::sra_i32_4(sc.cos_vals, 15));
1147 }
1148
1150 FASTLED_FORCE_INLINE s16x16x4 sin() const {
1151 s16x16x4 sin_out, cos_out;
1152 sincos(sin_out, cos_out);
1153 return sin_out;
1154 }
1155
1157 FASTLED_FORCE_INLINE s16x16x4 cos() const {
1158 s16x16x4 sin_out, cos_out;
1159 sincos(sin_out, cos_out);
1160 return cos_out;
1161 }
1162};
1163
1164// Include simd_ops.h to implement cross-type operations
1165// Must come after all types are defined
1166#include "fl/math/fixed_point/simd_ops.h" // allow-include-after-namespace
1167
1168} // namespace fl
Alignment macros and utilities for FastLED.
platforms::simd_u32x4 simd_u32x4
Definition types.h:26
constexpr int type_rank< T >::value
FASTLED_FORCE_INLINE SinCos32_simd sincos32_simd(simd::simd_u32x4 angles) FL_NOEXCEPT
Process 4 angles simultaneously, returning vectorized sin/cos values SIMD-optimized: vectorized angle...
Definition sin32.h:145
expected< T, E > result
Alias for expected (Rust-style naming)
Definition result.h:31
Base definition for an LED controller.
Definition crgb.hpp:179
#define FASTLED_FORCE_INLINE
Umbrella header for SIMD subsystem.
Cross-type SIMD fixed-point operations (implemented after all types are defined)
4-wide s0x32 vector (normalized values [-1, 1]) Backed by 128-bit SIMD register (4× i32 in Q31 format...
Definition s0x32x4.h:17
FASTLED_FORCE_INLINE s16x16x4 min(s16x16x4 b) const
Element-wise minimum.
Definition s16x16x4.h:90
FASTLED_FORCE_INLINE s16x16x4 clamp(s16x16x4 lo, s16x16x4 hi) const
Clamp to [lo, hi].
Definition s16x16x4.h:100
FASTLED_FORCE_INLINE s16x16x4 cos() const
Compute cosine of 4 angles (in radians)
Definition s16x16x4.h:138
FASTLED_FORCE_INLINE s16x16x4 operator-() const
Definition s16x16x4.h:60
FASTLED_FORCE_INLINE s16x16x4 operator*(s16x16x4 b) const
Definition s16x16x4.h:55
FASTLED_FORCE_INLINE void store(s16x16 *ptr) const
Definition s16x16x4.h:36
FASTLED_FORCE_INLINE void sincos(s16x16x4 &out_sin, s16x16x4 &out_cos) const
Compute sin and cos of 4 angles (in radians) Results written to out_sin and out_cos.
Definition s16x16x4.h:114
FASTLED_FORCE_INLINE s16x16x4 sin() const
Compute sine of 4 angles (in radians)
Definition s16x16x4.h:131
FASTLED_FORCE_INLINE s16x16x4 abs() const
Absolute value: branchless via mask and xor.
Definition s16x16x4.h:81
FASTLED_FORCE_INLINE s16x16x4 operator>>(int shift) const
Definition s16x16x4.h:66
static FASTLED_FORCE_INLINE s16x16x4 from_raw(simd::simd_u32x4 r)
Definition s16x16x4.h:24
FASTLED_FORCE_INLINE s16x16x4 operator<<(int shift) const
Definition s16x16x4.h:70
simd::simd_u32x4 raw
Definition s16x16x4.h:20
FASTLED_FORCE_INLINE s16x16x4 max(s16x16x4 b) const
Element-wise maximum.
Definition s16x16x4.h:95
static FASTLED_FORCE_INLINE s16x16x4 load(const s16x16 *ptr)
Definition s16x16x4.h:31
static FASTLED_FORCE_INLINE s16x16x4 set1(s16x16 value)
Definition s16x16x4.h:41
FASTLED_FORCE_INLINE s16x16x4 lerp(s16x16x4 b, s16x16 t) const
Linear interpolation: a + (b - a) * t (using Q16 multiply) t: s16x16 interpolation factor in [0,...
Definition s16x16x4.h:106
FASTLED_FORCE_INLINE s16x16x4 operator+(s16x16x4 b) const
Definition s16x16x4.h:47
4-wide s16x16 vector (general fixed-point) Backed by 128-bit SIMD register (4× i32 in Q16 format)
Definition s16x16x4.h:19