IQMath to Float compile error

Tim11828

Hi,

I'm getting a compile error with the following line of code when changing from IQMath to floating point (done in switch in IQMathLib.h):

clarke1.As=_IQ15toIQ((AdcRegs.ADCRESULT0>>1)-_IQ15(0.55))<<1;

"2xPM_Motors.c", line 656: error: expression must have integral type

Since I'm doing floating point, I could rewrite the code to simple floating point, but I'm curious as to why this no longer compiles? (It appears the last left shift is an issue.)

Thanks.

over 15 years ago

0 Lori Heustess over 15 years ago

TI__Guru* 91990 points

Tim,

A shift operation moves bits left or right. Its an easy way to scale an integer by 2, 4, or divide by 1/2, 1/4 etc.

For an integer this makes sense because of the format of the number. Floating-point numbers, however, have quite a different format. It has a sign bit, an exponent field and a mantissa. Each of these fields is a fixed number of bits and the corresponding normal range of values is given by an equation like this: (-1)^signbits ×2^{exponentbits-127}× 1.Mantissa. If you were to simply shift bits one way or the other it would not give you the results you expect.

In your project you could do something like this - it would still allow you to quickly flip between float and IQ

//
// Example:
// Convert from float to IQ15
 
//
// If MATH_TYPE == IQ_MATH
// Use the IQmath conversion function
//
 
#if MATH_TYPE == IQ_MATH
   PwmReg = (int16)_IQtoIQ15(Var1);
 
//
// If MATH_TYPE == FLOAT_MATH
// Scale by 2^15 = 32768.0
//
 
#else // MATH_TYPE is FLOAT_MATH
   PwmReg = (int16)(32768.0L*Var1);
#endif

Hope that helps

-Lori

0 Tim11828 over 15 years ago in reply to Lori Heustess

Expert 1845 points

Hi Lori,

I had somehow forgot that I was now working in floating point, and so that IS an invalid operation.

Thanks for the reminder.

Tim

C2000™︎ microcontrollers

C2000 microcontrollers forum

IQMath to Float compile error