CCS/TMS320F28379D: Float to int conversion with the CLA

Ralf Bartling

Expert 1155 points

Part Number: TMS320F28379D

Tool/software: Code Composer Studio

Hello,

I have a problem with the conversion from float to int with the CLA. The following picture shows the code.

The problem is the following line.

DaccRegs.DACVALS.all = (short)(u_drossel*(4095/(3.3/0.5*20))) + 2048;

u_drossel is a local float variable and contains a value of -60.0. The expression in brackets (4095/(3.3/0.5*20)) is equal to 31.022. This value is multiplicated with u_drossel and should be converted to a signed short variable. But the compiler uses MF32TOUI16 mnemonic instead of MF32TOI16, so the negative range of the value is not realized.

How can I fix this problem?

Thank you very much.

Ralf

over 5 years ago

0 Lori Heustess over 5 years ago

TI__Guru* 93270 points

Ignore this post - it was not correct.

Lori

0 Lori Heustess over 5 years ago in reply to Lori Heustess

TI__Guru* 93270 points

Ralf - I realized after review with the DAC expert that my understanding of the problem statement was not correct. Please disregard my previous response. I am looking into this.

0 Lori Heustess over 5 years ago in reply to Lori Heustess

TI__Guru* 93270 points

Ralf,

If the addition is performed before the cast, it looks correct.

i.e. try this:

DaccRegs.DACVALS.all = (short)( (u_drossel*(4095/(3.3/0.5*20))) + 2048 );

I'm consulting with the compiler team to understand why it isn't working the original way (i.e. cast before addition).

Lori

0 Lori Heustess over 5 years ago in reply to Lori Heustess

TI__Guru* 93270 points

Ralf,

It looks like this is a bug. I have filed a defect report for the compiler. In the meantime the customer can use the workaround I posted above.

Regards

Lori

C2000™︎ microcontrollers

C2000 microcontrollers forum

CCS/TMS320F28379D: Float to int conversion with the CLA