Compiler/LAUNCHXL-F28069M: Problem with conversion from Uint16 to float

Dear All

     Do you need a sample project to verify?

Let me know and I will provide a link in a private message

We need a test case we can build.  It does not have to run.  I am confident inspecting the compiler generated assembly code will be good enough to work out what happened.  For the function which contains this code ...

user4957458 said:

while(1)
{
AdcConversion_run(conv);
phase1_current=(((float)(conv[0]-conv[3]))/4096.0f)*phase_1_gain; 
phase3_current=(((float)(conv[1]-conv[4]))/4096.0f)*phase_3_gain; 
phase2_current=(((float)(conv[2]-conv[5]))/4096.0f)*phase_2_gain; 
}

... please follow the directions in the article How to Submit a Compiler Test Case.

Thanks and regards,

-George

I cannot reproduce the problem with C2000 16.9.6.LTS with the options you provided; as far as I can tell, the code is correct.  I've enclosed a cut-down test case demonstrating that the compiler does the right thing.  You say that you have observed "(float)(conv[0]-conv[3]) is equal to 0.0".  How were you able to do this?  Were you using a printf statement or the CCS watch window?

/*
  -v28
  -ml
  -mt
  --cla_support=cla0
  --float_support=fpu32
  --tmu_support=tmu0
  --vcu_support=vcu0
  -Ooff
  --opt_for_speed=5
  --fp_mode=relaxed
  --advice:performance=all
  -g
  --c99
  --diag_warning=225
  --diag_wrap=off
  --display_error_number
 */

#include <stdio.h>
#include <stdint.h>

float phase_1_gain=3.3f;
float phase_2_gain=3.3f;
float phase_3_gain=3.3f;

void try_signed()
{
    int16_t conv[6] = { 1500, 1500, 1500, 0, 0, 0 };

    float phase1_current=(((float)(conv[0]-conv[3]))/4096.0f)*phase_1_gain;
    float phase3_current=(((float)(conv[1]-conv[4]))/4096.0f)*phase_3_gain;
    float phase2_current=(((float)(conv[2]-conv[5]))/4096.0f)*phase_2_gain;

    printf("p1c: %f\n", phase1_current);
    printf("p2c: %f\n", phase2_current);
    printf("p3c: %f\n", phase3_current);
}

void try_unsigned()
{
    uint16_t conv[6] = { 1500, 1500, 1500, 0, 0, 0 };

    float phase1_current=(((float)(conv[0]-conv[3]))/4096.0f)*phase_1_gain;
    float phase3_current=(((float)(conv[1]-conv[4]))/4096.0f)*phase_3_gain;
    float phase2_current=(((float)(conv[2]-conv[5]))/4096.0f)*phase_2_gain;

    printf("p1c: %f\n", phase1_current);
    printf("p2c: %f\n", phase2_current);
    printf("p3c: %f\n", phase3_current);
}

void main()
{
    try_signed();
    try_unsigned();
}

In mode Release works. I tested with a "if ... then" with the leds