OMAP-L138: Floating point support for ARM using SYS/BIOS

Dieter Vogel

Intellectual 390 points

Part Number: OMAP-L138
Other Parts Discussed in Thread: SYSBIOS, OMAPL138

I am using the ARM of the OMAP-L138

Code Composer Studio 5.5

I am using SYS/BIOS

Target processor version is set to 5e

Designate code state is set to 32

Floating point support is set to VFPv2

When I change floating point support from vfplib to VFPv2, I get the error:

Tag_ABI_VFP_args attribute value of "0" that is different than one previously seen ("1"); combining incompatible files

If I remove some of the functions that use floating point numbers, I do not get this error. I read some of the previous posts about this error. Most of them talk about using a library that was not compiled with the correct flags. I am not using any libraries like that.

If I try to use --diag_suppress=16004 then my code crashes and says “ti.sysbios.family.arm.exc.Exception: line 180: E_undefinedInstruction:”

Is there something else that I need to do to use the floating point coprocessor of this ARM 9E? Thanks.

over 7 years ago

0 Cvetolin Shulev-XID over 7 years ago

TI__Guru 65405 points

The team is notified. They will post their feedback directly here.

BR
Tsvetolin Shulev

0 judahvang over 7 years ago

TI__Mastermind 32475 points

Hi,

If you are using BIOS than you are using libraries that are built differently. There might be other libraries like the rts and xdc runtime too.
I'm sure that BIOS is not built with the VFPv2 option. What you can try to determine this is:

In your .cfg file, you can do:
print(BIOS.customCCOpts)

This will print out the compiler opts used for building the BIOS libraries.
You can then either replace an exsiting line or add on to it with the following:

BIOS.customCCOpts = BIOS.customCCOpts.replace(“blah”, “blah”);

Or

BIOS.customCCOpts += “blah”;

Now, I'm not sure if this is enough as there could be other libraries not part of BIOS that needs to be rebuilt with the same compile options, but its a start.

0 Karl Wechsler over 7 years ago in reply to judahvang

TI__Mastermind 20805 points

Do you have an external floating point coprocessor on your board? From my read of the docs, it doesn't look like there is a hardware floating point support on-chip for to the Arm9 device on OMAPL13x. The compiler is generating an op-code that the device cannot handle, hence the illegal instruction exception.

0 Dieter Vogel over 7 years ago in reply to judahvang

Intellectual 390 points

Thank you, I will try that. Is there a difference between using vdplib or VFPv2? The ARM Optimizing C/C++ Compiler v18.1.0.LTS users guide suggests using VFPv2 for the ARM9E.

0 judahvang over 7 years ago in reply to Dieter Vogel

TI__Mastermind 32475 points

Yes, there is a difference. In the first case it uses a software library for doing floating point math. In the second case (VFPv2) it uses hardware for doing floating point math but your device must have hardware support for this. I think some ARM9 devices do not have floating point hardware.

0 Karl Wechsler over 7 years ago in reply to judahvang

TI__Mastermind 20805 points

I think that this thread can be closed. There is no hardware floating support on the Arm9 on OMAPL138. You must use the software floating point library. The hardware library is not supported.

0 Mukul Bhatnagar over 7 years ago in reply to Karl Wechsler

TI__Guru* 83825 points

Here is the reference to an old post that confirms that ARM9 on OMAPL1x family do not support hardware floating point

e2e.ti.com/.../491409

Processors

Processors forum

OMAP-L138: Floating point support for ARM using SYS/BIOS