TMS320F28379D: CLA

? ?

Part Number: TMS320F28379D

Tool/software:

Hello Ti engineers!

I want to what is the difference between using "\" and "CLAdiv function" in the .CLA code. I found both of "\" and "CLAdiv" can do the division, but the "\" method seems work faster.

Another question if the "\" can work well, why we still need the "CLAdiv"?

Thanks.

5 months ago

0 Ozino Odharo 5 months ago

TI__Mastermind 21335 points

Hi,

Can you confirm if you are using the CLA core in your application. If so, The TMS320C28x Control Law Accelerator math library is a collection of optimized floating-point math functions for controllers with the CLA. This source code library includes several C callable assembly math functions. The library is meant to work with the CLA C compiler.

Control Law Accelerator (CLA), a hardware coprocessor that performs math-intensive calculations in parallel with the main CPU. This library provides optimized, single-precision, floating-point math functions that are specifically tailored for the CLA architecture.

This document more details about the library.

/cfs-file/__key/communityserver-discussions-components-files/171/8345.CLA_2D00_MATH_2D00_LIB_2D00_UG.pdf

Regards,

Ozino

0 ? ? 5 months ago in reply to Ozino Odharo

Prodigy 25 points

Hi,

I have added the CLA math lib in the project. In the map file, I see the CLAdiv.obj. If I remove the CLAdiv in the .cla file, will not see the CLAdiv.obj in map file, but the code still work well.

I'm puzzled as to why " \" can be used in the CLA.

Regards,

+1 Ozino Odharo 5 months ago in reply to ? ?

TI__Mastermind 21335 points

Hi,

You should be able to use the divide option on the CLA. Though there is no support for integer division.

The performance should be the same between the two functions. See this post: TMS320F28377D-EP: CLA Division Operator "/" vs CLAdiv function

below is a summary of fundamental math operations not supported on the CLA. https://dev.ti.com/tirex/explore/node?node=A__ASh.QBtmaD.DbEzgWFwnEw__C28X-ACADEMY__1sbHxUB__LATEST

No support for certain fundamental math operations
- integer division: z = x/y;
- modulus (remainder): z = x%y;
- unsigned 32-bit integer compares
CLA Architecture is designed for 32-bit data types
- 16-bit computations incur overhead for sign-extension
- 16-bit values mostly used to read/write 16-bit peripheral registers
- There is no SW or HW support for 64-bit integer or floating point

Note: the CLA has instructions that are a subset of the C28 + FPU instructions. This forum post highlights the performance upsides to using the CLA vs C28.

TMS320F28075: Writing more efficient c code on CLA

This link also talks more about the comparison to CLA: https://software-dl.ti.com/C2000/docs/cla_software_dev_guide/cla_fpu_comparison.html

Regards,

Ozino

C2000™︎ microcontrollers

C2000 microcontrollers forum

TMS320F28379D: CLA