Detailed Description

Linear interpolation algorithms.

See test_interpolation_3d.cpp

Date: Oct 17, 2013

Author: Andrey Belomutskiy, (c) 2012-2020; Dmitry Sidin, (c) 2015

Functions
static void	testBinary ()

float	interpolateMsg (const char *msg, float x1, float y1, float x2, float y2, float x)
	Linear interpolation by two points. More...

float	interpolateClampedWithValidation (float x1, float y1, float x2, float y2, float x)

float	interpolateClamped (float x1, float y1, float x2, float y2, float x)

int	findIndex2 (const float array[], unsigned size, float value)

int	findIndex (const float array[], int size, float value)

void	initInterpolation ()

Variables
float	array16 [] = { 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16 }

Function Documentation

◆ findIndex()

int findIndex	(	const float	array[],
		int	size,
		float	value
	)

Definition at line 155 of file interpolation.cpp.

                                                           {
     return findIndexMsg("", array, size, value);
 }

Referenced by testBinary().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ findIndex2()

int findIndex2	(	const float	array[],
		unsigned	size,
		float	value
	)

Another implementation, which one is faster?

it should be "if (j < size && array[j] <= value)" but in our case size is always power of 2 thus size is always more then j

Definition at line 131 of file interpolation.cpp.

                                                                 {
     efiAssert(ObdCode::CUSTOM_ERR_ASSERT, !cisnan(value), "NaN in findIndex2", 0);
     efiAssert(ObdCode::CUSTOM_ERR_ASSERT, size > 1, "size in findIndex", 0);
 //  if (size <= 1)
 //      return size && *array <= value ? 0 : -1;
  
     signed i = 0;
     //unsigned b = 1 << int(log(float(size) - 1) / 0.69314718055994530942);
     unsigned b = size >> 1; // in our case size is always a power of 2
     efiAssert(ObdCode::CUSTOM_ERR_ASSERT, b + b == size, "Size not power of 2", -1);
     for (; b; b >>= 1) {
         unsigned j = i | b;
         /**
          * it should be
          * "if (j < size && array[j] <= value)"
          * but in our case size is always power of 2 thus size is always more then j
          */
         // efiAssert(ObdCode::CUSTOM_ERR_ASSERT, j < size, "size", 0);
         if (array[j] <= value)
             i = j;
     }
     return i || *array <= value ? i : -1;
 }

Referenced by testBinary().

Here is the caller graph for this function:

◆ initInterpolation()

void initInterpolation ( )

Definition at line 159 of file interpolation.cpp.

                          {
 #if BINARY_PERF && ! EFI_UNIT_TEST
     addConsoleAction("binarytest", testBinary);
 #endif
 }

Referenced by commonInitEngineController().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ interpolateClamped()

float interpolateClamped	(	float	x1,
		float	y1,
		float	x2,
		float	y2,
		float	x
	)

todo: use 'interpolateClampedWithValidation' wider?

See also: interpolateMsg

Definition at line 115 of file interpolation.cpp.

                                                                           {
     // note how we assume order of x1 and x2 here! see also interpolateClampedWithValidation
     if (x <= x1)
         return y1;
     if (x >= x2)
         return y2;
  
     // todo: do we care with code duplication with interpolateMsg above?
     float a = INTERPOLATION_A(x1, y1, x2, y2);
     float b = y1 - a * x1;
     return a * x + b;
 }

Referenced by ThrottleModelBase::estimateThrottleFlow(), flexCallback(), IdleController::getClosedLoop(), getCrankingAdvance(), getCrankingFuel3(), IdleController::getIdleTimingAdjustment(), IdleController::getOpenLoop(), getRunningAdvance(), IdleController::getRunningOpenLoop(), EtbController::getSetpointEtb(), FuelComputer::getStoichiometricRatio(), IFuelComputer::getTCharge(), IFuelComputer::getTChargeCoefficient(), AirmassVeModelBase::getVe(), interpolateClampedWithValidation(), HellaOilLevelSensor::onEdge(), setHysteresis(), LimpManager::updateRevLimit(), and updateVrThresholdPwm().

Here is the caller graph for this function:

◆ interpolateClampedWithValidation()

float interpolateClampedWithValidation	(	float	x1,
		float	y1,
		float	x2,
		float	y2,
		float	x
	)

Definition at line 104 of file interpolation.cpp.

                                                                                         {
     if (x1 >= x2) {
         criticalError("interpolateClamped %f has to be smaller than %f", x1, x2);
     }
     return interpolateClamped(x1, y1, x2, y2, x);
 }

Here is the call graph for this function:

◆ interpolateMsg()

float interpolateMsg	(	const char *	msg,
		float	x1,
		float	y1,
		float	x2,
		float	y2,
		float	x
	)

Linear interpolation by two points.

Parameters

x1	key of the first point
y1	value of the first point
x2	key of the second point
y2	value of the second point
X	key to be interpolated

Note: For example, "interpolateMsg("", engineConfiguration.tpsMin, 0, engineConfiguration.tpsMax, 100, adc);"

See also: interpolateClamped

we could end up here for example while resetting bins while changing engine type

Definition at line 74 of file interpolation.cpp.

                                                                                        {
     if (cisnan(x1) || cisnan(x2) || cisnan(y1) || cisnan(y2)) {
         warning(ObdCode::CUSTOM_ERR_INTERPOLATE_1, "interpolate%s: why param", msg);
         return NAN;
     }
     if (cisnan(x)) {
         warning(ObdCode::CUSTOM_ERR_INTERPOLATE_2, "interpolate%s: why X", msg);
         return NAN;
     }
     // todo: double comparison using EPS
     if (x1 == x2) {
         /**
          * we could end up here for example while resetting bins while changing engine type
          */
         warning(ObdCode::CUSTOM_ERR_INTERPOLATE_3, "interpolate%s: Same x1 and x2 in interpolate: %.2f/%.2f", msg, x1, x2);
         return NAN;
     }
  
     // a*x1 + b = y1
     // a*x2 + b = y2
 //  efiAssertVoid(ObdCode::CUSTOM_ERR_ASSERT_VOID, x1 != x2, "no way we can interpolate");
     float a = INTERPOLATION_A(x1, y1, x2, y2);
     if (cisnan(a)) {
         warning(ObdCode::CUSTOM_ERR_INTERPOLATE_4, "interpolate%s: why a", msg);
         return NAN;
     }
     float b = y1 - a * x1;
     return a * x + b;
 }

Referenced by decodeTpsSentValue(), getAdvanceForRpm(), getAfr(), IFuelComputer::getTChargeCoefficient(), lua_interpolate(), and setLinearCurve().

Here is the call graph for this function:

Here is the caller graph for this function:

◆ testBinary()

static void testBinary ( )

static

Definition at line 25 of file interpolation.cpp.

                          {
     const int size16 = 16;
  
     uint32_t totalOld = 0;
     uint32_t totalNew = 0;
  
     for (int v = 0; v <= 16; v++) {
         uint32_t timeOld;
         {
             uint32_t start = getTimeNowLowerNt();
             int temp = 0;
             for (int i = 0; i < COUNT; i++) {
                 temp += findIndex(array16, size16, v);
             }
             timeOld = getTimeNowLowerNt() - start;
         }
         uint32_t timeNew;
         {
             uint32_t start = getTimeNowLowerNt();
             int temp = 0;
             for (int i = 0; i < COUNT; i++) {
                 temp += findIndex2(array16, size16, v);
             }
             timeNew = getTimeNowLowerNt() - start;
         }
         efiPrintf("for v=%d old=%d ticks", v, timeOld);
         efiPrintf("for v=%d new=%d ticks", v, timeNew);
  
         totalOld += timeOld;
         totalNew += timeNew;
     }
     efiPrintf("totalOld=%d ticks", totalOld);
     efiPrintf("totalNew=%d ticks", totalNew);
  
 }