Add configuration file

Fusion/FusionConfig.h

@@ -0,0 +1,28 @@
+/**
+ * @file FusionConfig.h
+ * @author Seb Madgwick
+ * @brief Library configuration.
+ */
+
+#ifndef FUSION_CONFIG_H
+#define FUSION_CONFIG_H
+
+//------------------------------------------------------------------------------
+// Definitions
+
+/**
+ * @brief Uncomment this definition or add as a preprocessor definition to use
+ * normal square root operations.
+ */
+//#define FUSION_USE_NORMAL_SQRT
+
+/**
+ * @brief Uncomment this definition or add as a preprocessor definition to
+ * disable inline functions.
+ */
+//#define FUSION_NO_INLINE
+
+#endif
+
+//------------------------------------------------------------------------------
+// End of file

Fusion/FusionInline.h

@@ -7,10 +7,17 @@
 #ifndef FUSION_INLINE_H
 #define FUSION_INLINE_H
 
+//------------------------------------------------------------------------------
+// Includes
+
+#include "FusionConfig.h"
+
 //------------------------------------------------------------------------------
 // Definitions
 
-#if defined _MSC_VER
+#if defined FUSION_NO_INLINE
+#define FUSION_INLINE inline
+#elif defined _MSC_VER
 #define FUSION_INLINE __forceinline
 #elif defined __GNUC__
 #define FUSION_INLINE inline __attribute__((always_inline))

Fusion/FusionMath.h

@@ -10,6 +10,7 @@
 //------------------------------------------------------------------------------
 // Includes
 
+#include "FusionConfig.h"
 #include "FusionInline.h"
 #include <math.h>
 #include <stdbool.h>
@@ -111,12 +112,6 @@ typedef union {
 #define M_PI (3.14159265358979323846)
 #endif
 
-/**
- * @brief Include this definition or add as a preprocessor definition to use
- * normal square root operations.
- */
-//#define FUSION_USE_NORMAL_SQRT
-
 //------------------------------------------------------------------------------
 // Inline functions - Degrees and radians conversion
 

README.md

@@ -139,6 +139,12 @@ where:
 
 Sensors are often mounted such that their axes are not aligned with the body axes. The library provides the `FusionRemap` function to remap sensor axes to the body frame using one of the 24 possible orthogonal axis permutations defined by `FusionRemapAlignment`.
 
-## Fast inverse square root
+## Optimisations
 
-Fusion uses [Pizer's implementation](https://pizer.wordpress.com/2008/10/12/fast-inverse-square-root/) of the [fast inverse square root](https://en.wikipedia.org/wiki/Fast_inverse_square_root) algorithm for vector and quaternion normalisation. Including the definition `FUSION_USE_NORMAL_SQRT` in [FusionMath.h](https://github.com/xioTechnologies/Fusion/blob/main/Fusion/FusionMath.h) or adding this as a preprocessor definition will use normal square root operations for all normalisation calculations. This will slow down execution speed for a small increase in accuracy. The increase in accuracy will typically be too small to observe on any practical system.
+### Inline functions
+
+The `FUSION_INLINE` macro implements either `__forceinline` or `inline __attribute__((always_inline))` to inline functions and optimise for speed at the cost of program size. This can be disabled by uncommenting the `FUSION_NO_INLINE` in [FusionConfig.h](https://github.com/xioTechnologies/Fusion/blob/main/Fusion/FusionConfig.h) or adding this as a preprocessor definition.
+
+### Fast inverse square root
+
+Vector and quaternion normalisation operations use [Pizer's implementation](https://pizer.wordpress.com/2008/10/12/fast-inverse-square-root/) of the [fast inverse square root](https://en.wikipedia.org/wiki/Fast_inverse_square_root) algorithm. This increases speed for devices without an FPU, at a negligible cost to accuracy. Uncommenting `FUSION_USE_NORMAL_SQRT` in [FusionConfig.h](https://github.com/xioTechnologies/Fusion/blob/main/Fusion/FusionConfig.h) or adding this as a preprocessor definition will use normal square root for all operations.