CCS5 Floating Point Support

Brian,

To date CCS does not have a Cortex A8 simulator.

Regarding the issues with the argument stack, would you mind sending a testcase that shows the issue? Since you are seeing this in the disassembly, I imagine the compiler may be at fault but this would require further investigation (the testcase goes a long way to help with this).

The registers can be added to the Expressions view by using their names starting with upper caps (check the attached screen).

Regards,

Rafael

Rafael

Thank you for your reply. I will check out adding registers using Expressions.   Below is a function call:

Please see the attached Word doc.  BTW:  This happens with the optimizer turned on.

The first Picture shows the following:

       SimpleInterpolateLSF(syntdenum, weightdenum,
           lsf, lsfdeq, iLBCenc_inst->lsfold,
           iLBCenc_inst->lsfdeqold,10, iLBCenc_inst);

The number 10 is the 3rd item on the stack  at R13 + 8.

The second picture  is after the function is stepped into.   R4 gets loaded with R13 + 0x80.  For some reason R4 is used as length in the following 2 calculations:

MOV R12, R2, R4, LSL #2  it should be MOV R12,R2,R12, #2

Hopefully this is clear.  Again, thank you for your help.

Do you know when A8 will be supported?

Brian

Brian,

Sorry, what I meant by testcase was a small project that could be compiled here at TI, so we would be able to experiment with the code and better investigate this. I can try to create a small project that has multiple parameters to a function call, but I cannot be sure if the issue will pop up.

If you are willing to send your complete project (the quickest way to see this issue and start working on it), then we could talk in private via friendship request.

Let me know what do you prefer, ok?

Cheers,

Rafael

Rafael

I'd be happy to put a test case together but you know how that works, the compiler may behave properly.  In any event I will try.

Thanks Again

Brian

Rafael

 I can display the registers in the Expressions window but there doesn't seem to be a number format for floating point.  Do I need to cast or something?

Thanks,

Brian

Brian,

Yes, you can typecast the register value. Check the attached screen.

I was experimenting here with some test code I had but still couldn't see any issues. What exactly are the compiler options used?

Regards,

Rafael

Rafael

Thanks for the info!

I attached a word doc with 2 compiler windows and a linker window.  If you need more info please let me know.

Best Regards

Brian

Hi Rafael

I have created a small project that just floats an integer counter and loads into 2 floating point arrays. As you can see from the Expressions window it doesn't seem to work.  Also  I don't know how to view the extended single precision registers.  I'll send the project in the next post because I can't seem to attach 2 files in one post.

Thanks!

Brian

Here is the project.