Unnecessary doubles in Accelerometer & Compass

[microbit-carlos]

In codal-core Accelerometer::recalculatePitchRoll():

• codal-core/source/driver-models/Accelerometer.cpp#L488

void Accelerometer::recalculatePitchRoll()
{
    double x = (double) sample.x;
    double y = (double) sample.y;
    double z = (double) sample.z;

    roll = atan2(x, -z);
    pitch = atan2(y, (x*sin(roll) - z*cos(roll)));

    // Handle to the two "negative quadrants", such that we get an output in the +/- 18- degree range.
    // This ensures that the pitch values are consistent with the roll values.
    if (z > 0.0)
    {
        double reference = pitch > 0.0 ? (PI / 2.0) : (-PI / 2.0);
        pitch = reference + (reference - pitch);
    }

    status |= ACCELEROMETER_IMU_DATA_VALID;
}

And Compass::getFieldStrength():

• codal-core/source/driver-models/Compass.cpp#L124

int Compass::getFieldStrength()
{
    Sample3D s = getSample();

    double x = s.x;
    double y = s.y;
    double z = s.z;

    return (int) sqrt(x*x + y*y + z*z);
}

@finneyj we probably don't need the extra precision and can downgrade them to floats?

[finneyj]

Thanks @microbit-carlos. Yes, I think this looks pretty safe!

I recall on v1 we had some rounding errors in the old compass calibration code that were causing some problems - this was probably a leftover from that...

[finneyj]

p.s. Can you tell how much RAM/FLASH we save from this?