fix #28, prevent promotion of constants to double

CHANGELOG.md

@@ -6,15 +6,22 @@ The format is based on [Keep a Changelog](http://keepachangelog.com/)
 and this project adheres to [Semantic Versioning](http://semver.org/).
 
 
+## [0.3.3] - 2025-03-23
+- fix #28, prevent promotion of constants to double. Kudos to Harrison3000
+- replace M_PI by a const float _PI_ to prevent promotion.
+- update examples
+- update readme.md (minor)
+- minor edits
+
+
 ## [0.3.3] - 2023-11-02
 - update readme.md
 - update keywords.txt
 - update changelog.md
 
-
 ## [0.3.2] - 2022-12-20
 - fix changelog
-- fix #25 change reference parameters to pointer 
+- fix #25 change reference parameters to pointer
   - isincos(float f, float \*si, float \*co)
   - isincos256(uint32_t v, int \*si, int \*co);
   - fix examples
@@ -35,9 +42,9 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 ----
 
 ## [0.2.1] - 2022-12-05
-- add atanFast(x) => faster and less accurate than atan(). 
+- add atanFast(x) => faster and less accurate than atan().
   Input range (-1..1) is fastest.
-- add atan2Fast(y, x) => faster and less accurate. 
+- add atan2Fast(y, x) => faster and less accurate.
   calls atanFast() + offset.
 - add example to measure performance atanFast
 - update readme.md
@@ -60,11 +67,11 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 - no functional changes
 
 ## [0.1.10] - 2022-04-15
-- fix #12 
+- fix #12
 - split .h in .h and .cpp Needed in case of more complex projects.
 
-## [0.1.9] - 2021-12-18  
-- update Arduino-CI, badges, 
+## [0.1.9] - 2021-12-18
+- update Arduino-CI, badges,
 - update library.json
 - minor edits
 
@@ -83,14 +90,14 @@ and this project adheres to [Semantic Versioning](http://semver.org/).
 - new table
 - added iasin() and iacos()
 
-In (0.1.4) an error was found in the optimize algorithm, so for 0.1.5 
-it was ran again and accuracy improved for **isin()** and **icos()**. 
-However **itan()** lost a (smaller) bit. 
+In (0.1.4) an error was found in the optimize algorithm, so for 0.1.5
+it was ran again and accuracy improved for **isin()** and **icos()**.
+However **itan()** lost a (smaller) bit.
 The gain outweighs the loss and so new table is kept.
 
 Performance has not changed.
 
-An initial version of a reverse lookup for **iasin(val)** and **iacos(val)** 
+An initial version of a reverse lookup for **iasin(val)** and **iacos(val)**
 is added, as it uses the same **isintable16\[\]** interpolation table.
 
 There is no **atan()** or **atan2()** replacement.
@@ -101,8 +108,8 @@ There is no **atan()** or **atan2()** replacement.
 - cleanup
 - examples
 
-The library (0.1.4) provides an **itan()** which improved accuracy 
-upon the (0.1.3) version and performance for the ESP32.  
+The library (0.1.4) provides an **itan()** which improved accuracy
+upon the (0.1.3) version and performance for the ESP32.
 Performance on AVR (UNO) is still an issue, accuracy is OK.
 
 
@@ -126,7 +133,7 @@ _eons passed_
 - added interpolation
 
 ## [0.1.01] - 2011-08-18
-- improved tables a bit 
+- improved tables a bit
 - changed param to float
 
 ## [0.1.00] - 2011-08-18

FastTrig.cpp

@@ -1,7 +1,7 @@
 //
 //    FILE: FastTrig.cpp
 //  AUTHOR: Rob Tillaart
-// VERSION: 0.3.3
+// VERSION: 0.3.4
 // PURPOSE: Arduino library for a faster approximation of sin() and cos()
 //    DATE: 2011-08-18
 //     URL: https://github.com/RobTillaart/FastTrig
@@ -11,6 +11,10 @@
 #include "FastTrig.h"
 
 
+const float _PI_  = 3.14159265;
+const float _PI_DIV_2_ = _PI_ / 2;
+
+
 //  91 x 2 bytes ==> 182 bytes
 //  use 65535.0 as divider
 uint16_t sinTable16[] = {
@@ -45,7 +49,7 @@ uint16_t sinTable16[] = {
 */
 
 
-// use 255.0 as divider
+//  use 255.0 as divider
 uint8_t sinTable8[] = {
   0, 4, 9, 13, 18, 22, 27, 31, 35, 40, 44,
   49, 53, 57, 62, 66, 70, 75, 79, 83, 87,
@@ -62,8 +66,8 @@ uint8_t sinTable8[] = {
 
 ///////////////////////////////////////////////////////
 //
-// GONIO INT EXPERIMENTAL
-// works with only whole degrees.
+//  GONIO INT EXPERIMENTAL
+//  works with only whole degrees.
 //
 int isin256(uint32_t v)
 {
@@ -174,15 +178,15 @@ float isin(float f)
     }
   }
 
-  //  float value  improves ~4% on avg error  for ~60 bytes.
+  //  float value improves ~4% on average error for ~60 bytes.
   uint16_t value = sinTable16[y];
 
   //  interpolate if needed
   if (remain > 0)
   {
     value = value + ((sinTable16[y + 1] - value) / 8 * remain) / 32;   //  == * remain / 256
   }
-  float g = value * 0.0000152590219;  //  = / 65535.0
+  float g = value * 0.0000152590219f;  //  = / 65535.0
   if (negative) return -g;
   return g;
 }
@@ -235,15 +239,15 @@ void isincos(float f, float *si, float *co)
     cneg = !cneg;
   }
 
-  //  float value  improves ~4% on avg error  for ~60 bytes.
+  //  float value improves ~4% on average error for ~60 bytes.
   //  SIN
   uint16_t value = sinTable16[y];
   //  interpolate if needed
   if (remain > 0)
   {
     value = value + ((sinTable16[y + 1] - value) / 8 * remain) / 32;  //  == * remain / 256
   }
-  *si = value * 0.0000152590219;  //  = / 65535.0
+  *si = value * 0.0000152590219f;  //  = / 65535.0
   if (sneg) *si = - *si;
 
   //  COS
@@ -254,7 +258,7 @@ void isincos(float f, float *si, float *co)
     remain = 256 - remain;
     value = value + ((sinTable16[90-y] - value) / 8 * remain) / 32;  //  == * remain / 256
   }
-  *co = value * 0.0000152590219;  //  = / 65535.0
+  *co = value * 0.0000152590219f;  //  = / 65535.0
   if (cneg) *co = - *co;
 }
 
@@ -382,19 +386,19 @@ float iatan(float f)
 float atanFast(float x)
 {
   //  remove two test will limit the input range but makes it even faster.
-  if ( x > 1)  return ( M_PI / 2) - atanHelper(1.0 / x);
-  if ( x < -1) return (-M_PI / 2) - atanHelper(1.0 / x);
+  if ( x > 1)  return (_PI_DIV_2_) - atanHelper(1.0 / x);
+  if ( x < -1) return (-_PI_DIV_2_) - atanHelper(1.0 / x);
   return atanHelper(x);
 }
 
 
 inline float atanHelper(float x)
 {
   float x2 = x * x;
-  return (((0.079331 * x2) - 0.288679) * x2 + 0.995354) * x;
+  return (((0.079331f * x2) - 0.288679f) * x2 + 0.995354f) * x;
 
   //  an even more accurate alternative, less fast
-  //  return ((((-0.0389929 * x2) + 0.1462766) * x2 - 0.3211819) * x2 + 0.9992150) * x;
+  //  return ((((-0.0389929f * x2) + 0.1462766f) * x2 - 0.3211819f) * x2 + 0.9992150f) * x;
 }
 
 
@@ -407,29 +411,29 @@ float atan2Fast(float y, float x)
   {
     if (y >= 0)
     {
-      if (fabs(y) >= fabs(x)) return M_PI / 2 - atanFast(x / y);
+      if (fabs(y) >= fabs(x)) return _PI_DIV_2_ - atanFast(x / y);
       return atanFast(y / x);
     }
-    if (fabs(y) >= fabs(x)) return -M_PI / 2 - atanFast(x / y);
+    if (fabs(y) >= fabs(x)) return -_PI_DIV_2_ - atanFast(x / y);
     return atanFast(y / x);
   }
   else
   {
     if (y >= 0)
     {
-      if (fabs(y) >= fabs(x)) return M_PI / 2 - atanFast(x / y);
-      return M_PI + atanFast(y / x);
+      if (fabs(y) >= fabs(x)) return _PI_DIV_2_ - atanFast(x / y);
+      return _PI_ + atanFast(y / x);
     }
-    if (fabs(y) >= fabs(x)) return -M_PI / 2 - atanFast(x / y);
-    return -M_PI + atanFast(y / x);
+    if (fabs(y) >= fabs(x)) return -_PI_DIV_2_ - atanFast(x / y);
+    return -_PI_ + atanFast(y / x);
   }
 }
 
 
 ///////////////////////////////////////////////////////
 //
 //  HYPOT
-//  related but not strict gonio.
+//  related but not strict goniometry.
 //
 //  hypotFast()  formula for faster hypot() at the price of accuracy
 //  experimental!
@@ -442,7 +446,7 @@ float hypotFast(float x, float y)
     a = fabs(y);
     b = fabs(x);
   }
-  float z = 0.917981 * (b + a / 2);
+  float z = 0.917981f * (b + a / 2);
   if (z > b) return z;
   return b;
 }

FastTrig.h

@@ -2,7 +2,7 @@
 //
 //    FILE: FastTrig.h
 //  AUTHOR: Rob Tillaart
-// VERSION: 0.3.3
+// VERSION: 0.3.4
 // PURPOSE: Arduino library for a faster approximation of sin() and cos()
 //    DATE: 2011-08-18
 //     URL: https://github.com/RobTillaart/FastTrig
@@ -18,7 +18,8 @@
 #endif
 
 
-#define FAST_TRIG_LIB_VERSION             (F("0.3.3"))
+#define FAST_TRIG_LIB_VERSION             (F("0.3.4"))
+
 
 #ifdef __cplusplus
 extern "C"
@@ -74,7 +75,7 @@ float atan2Fast(float y, float x);
 ///////////////////////////////////////////////////////
 //
 //  HYPOT
-//  related but not strict gonio.
+//  related but not strict goniometry.
 //
 //  hypotFast()  formula for faster hypot() at the price of accuracy
 //  experimental!

LICENSE

@@ -1,6 +1,6 @@
 MIT License
 
-Copyright (c) 2011-2024 Rob Tillaart
+Copyright (c) 2011-2025 Rob Tillaart
 
 Permission is hereby granted, free of charge, to any person obtaining a copy
 of this software and associated documentation files (the "Software"), to deal

README.md

@@ -47,14 +47,20 @@ Of course this introduces an error but the error is small and performance is sti
 quite fast (which was the goal).
 
 
+### Related
+
+- https://github.com/RobTillaart/fast_math
+- https://github.com/RobTillaart?tab=repositories&q=math
+
+
 ## Interface
 
 ```cpp
 #include "FastTrig.h"
 ```
 
 
-#### Lookup tables
+### Lookup tables
 
 The lookup tables are optimized (sketch provided) to minimize the error when using the interpolation, 
 this implies that the points in the table might not be optimal when you use only whole degrees. 
@@ -71,10 +77,10 @@ however it is great to use in a LEDstrip or motor movements when less accuracy i
 
 Although the tables can be written to, it is advised not to do so.
 
-*OK, the optimize example does a write to improve the table to minimize errors*
+_OK, the optimize example does a write to improve the table to minimize errors_
 
 
-#### atan, atan2
+### atan, atan2
 
 Since version 0.2.1 two functions are added:
 - **float atanFast(float f)** input range -1 .. 1 is faster.
@@ -89,7 +95,7 @@ Use **fastTrig_atan_performance.ino** to check the gain on your board.
 Price is that the values are less accurate, but the difference is < 0.001.
 
 
-#### isin256, icos256, isincos256
+### isin256, icos256, isincos256
 
 Version 0.3.0 added these experimental functions:
 
@@ -102,7 +108,7 @@ returns both the sin(v)\*256 and the cos(v)\*256 of the same angle.
 Faster than both individual calls together.
 
 
-#### isincos
+### isincos
 
 Version 0.3.0 added this experimental function:
 
@@ -113,7 +119,7 @@ There is a minor difference between the value of the **float co** compared to **
 This need some investigation ( truncating ?)
 
 
-#### hypotFast
+### hypotFast
 
 Strictly **hypot()** is no gonio function but it is often used 
 for calculating length in polar coordinates.
@@ -367,6 +373,7 @@ See examples
 
 #### Should
 
+- clean example headers (consistency)
 - write more tests to verify values.
 - test performance on more platforms.
 - investigate the difference between **isincos()** and **icos()**.