Floating Point: MSP432 only 3x faster than MSP430?

ADAM ROTHROCK said:

The MSP432 should be much faster because of the FPU, right? Perhaps I don't have the FPU enabled correctly?

Any 64-bit (double) precision operations cause software routines to be called instead of using the hardware FPU.

In your example floating point constants such as 3.141 are as per the C standard treated as double precision, which causes the compiler to perform double precision calculation using software routines.

If you add a "f" suffix to such constants the constants become single precision and the compiler can then use the hardware FPU.

Looking at the generated assembler for your example showed the compiler calling software routines.

When if was changed to make the constants single precision then the compiler generated hardware FPU instructions instead:

    float ans;
    float op;

    while(1)
    {
    // Toggle P1.0 output
    GPIO_toggleOutputOnPin(
    GPIO_PORT_P1,
    GPIO_PIN0
    );

    ans = 0;
    //each loop should be 1000 cycles
    while(ans < 3141.0f)
    {
    //p2.7 is pulled low
    if(GPIO_getInputPinValue(GPIO_PORT_P2,GPIO_PIN7) == 0x80)
    {
    op = 0;
    }
    else
    op = 1.0f;
    ans = ans + 3.141f * op;
    }

    }

I haven't timed this updated example, but from looking at the difference in the generated assembler it should be faster [assuming it is acceptable for your results to be calculated entirely using single-precision].