SP Float to Int, rounding behavior

Steven Tuttle

In the C67x+ chips, the SPINT rounding behavior is controlled via rounding mode bits in the FPADCR register. Are these bits changed from the power-on defaults by any of the optimized DSP and/or math function libraries?

Are there any TI supplied libraries that change these rounding bits from their default state? If so, what are those library functions, and do they always restore these rounding mode bits after they are done?

Also, if there are any TI libraries that change these mode bits, then what happens to those mode bits when an interrupt occurs?

Thanks for your help. Steve

over 14 years ago

0 Brad Griffis over 14 years ago

TI__Guru*** 125430 points

It's hard to say with absolute certainty, but as far as I know none of the libraries change these bits.

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

Thanks for your reply.

Are you a TI employee? Do you have access to the source code for the libraries that are provided by TI?

If you do have access to that source code, how can you say this cannot be answered with absolute certainty?

This is an important question, the answer to which affects the accuracy of our computations.

If you can't answer it with certainty, where can I obtain the source code to these libraries, so I can verify it myself?

Thanks for your help.

Steve

0 Brad Griffis over 14 years ago in reply to Steven Tuttle

TI__Guru*** 125430 points

Steven Tuttle said:
Are you a TI employee?

Yes, if you view the posts on the forum you can see who is and is not a TI employee.

Steven Tuttle said:
If you can't answer it with certainty, where can I obtain the source code to these libraries, so I can verify it myself?

Please list the libraries of interest and I'll be happy to point you to the source code. I think the only library that I can not point you to source code is DSP/BIOS 5.x but in that case the API Guide specifically mentions that it does not deal with FADCR, FAUCR, and FMCR.

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

These are libraries in my project:

C672x\RomPatch\applypatch.obj
C672x\RomPatch\c672xrompatchv1_00_00.lib
C6000\cgtools\lib\rts6700.lib
c67xdsplibR.lib
fastrts67xR.lib

I have left out the CSL libs, since I already have the source code for those libraries.

Thanks for your help.

Steve

0 Brad Griffis over 14 years ago in reply to Steven Tuttle

TI__Guru*** 125430 points

Our libraries ship with a ".src" file in the lib directory. You can open that file in any text editor.

DSPLIB:http://focus.ti.com/docs/toolsw/folders/print/sprc121.html

FASTRTS: http://focus.ti.com/docs/toolsw/folders/print/sprc060.html

By the way, there is a 2.00 ROM patch that you should probably use too. The ROM patch is related to dMAX so that wouldn't have any affect here.

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

Thanks for the links.

Thanks also for the heads up on the Rom Patch revision. How does that patch affect dMAX operation?

When does the code in ROM get used for dMAX?

Steve

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

The link to sprc060.html takes me to a web page that has a 'Download' button, but the 'Download' button is a dead link:

http://focus.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=sprc060&fileType=zip

It just brings up a web page that says 'Sorry, we could not find that page'.

Steve

0 Brad Griffis over 14 years ago in reply to Steven Tuttle

TI__Guru*** 125430 points

It's a firmware update for the dMAX. I don't think I can say much more about the internal operation, though it would be fully independent of the stuff you're talking about now.

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

OK, I didn't realize that dMAX used firmware in the ROM - I understand that now.

I know you can't say much about the internals of the firmware, but can you at least describe the symptoms of the dMAX fault when one is using that v1.00 ROMpatch code, rather than the v2.00 ROMpatch? That would good to know for our own firmware management records.

Steve

0 Brad Griffis over 14 years ago in reply to Steven Tuttle

TI__Guru*** 125430 points

Steven Tuttle said:

Brad,

The link to sprc060.html takes me to a web page that has a 'Download' button, but the 'Download' button is a dead link:

http://focus.ti.com/general/docs/lit/getliterature.tsp?literatureNumber=sprc060&fileType=zip

It just brings up a web page that says 'Sorry, we could not find that page'.

Steve

Does this work:

http://www.ti.com/lit/zip/sprc060

You might try a few times. For some weird reason I once saw the error you described but it worked correctly several other times. ?

0 Steven Tuttle over 14 years ago in reply to Brad Griffis

Intellectual 820 points

Brad,

Yes, that link did work - Thanks!

Steve

0 Steven Tuttle over 14 years ago in reply to Steven Tuttle

Intellectual 820 points

I will repeat my question:

Steve

0 Gagan Maur over 14 years ago in reply to Steven Tuttle

TI__Expert 8150 points

Steve, no issues at all in sharing the details. The issue with the firmware image 1.00 was that the boot of dMAX wouldn't succeed with that. Thus, it was must to patch the ROM firmware image for the device at runtime. If more specific details help, we would be willing to share those as well. Please let me know.
Note the patched image also enables certain new transfer types that were not supported in the image that went into the ROM.
Also, note the dMAX firmware image is completely separate from C67x and has no association with the DSP.

Regards,
Gagan

0 Steven Tuttle over 14 years ago in reply to Gagan Maur

Intellectual 820 points

Gagan,

Thanks for the answer. So is the 'ROMpatch' just for dMAX firmware, or are there other parts of the C672x ROM that are patched as well?

Steve

0 Gagan Maur over 14 years ago in reply to Steven Tuttle

TI__Expert 8150 points

Steve, It is just for the dMAX firmware.
Best Regards,
Gagan

Processors

Processors forum

SP Float to Int, rounding behavior