Upgrade to SYS/BIOS 6.35.03.47 increased cpu load dramatically

Hi David --

I reviewed the bug database for changes between these releases and do not see anything that could explain this.

Are you only changing one variable (6.35.01 vs 6.35.03)?     Everything else is the same?    Have you looked at the 2 .map files to see if things are placed functionally differently?

How are you measuring CPU load?

Thanks,
-Karl- 

Hi Karl,

Thanks for your reply - sorry that mine has been rather slow.  Here are my answers to your questions.

Yes, we are only changing one variable - moving from SYS/BIOS 6.35.1.29 to 6.35.3.47.  Everything else is the same.  I saw my cpu load change from 16% to 43%.  I also got ~43% when I tried SYS/BIOS 6.35.4.50.

I have checked the map files and the differences look minor - changes of variable address of only 10's of bytes.

We are measuring cpu load by calling TI function LOAD_getcpuload().

Do you have any ideas about how to investigate what's going on please?

BR

David

Can you please try 6.35.02.45 as well?    This would help us narrow it down.

Thanks,
-Karl- 

You can get BIOS releases here but you probably already know that.

http://software-dl.ti.com/dsps/dsps_public_sw/sdo_sb/targetcontent/bios/sysbios/index.html

I really have no great ideas, but this will eliminate a few more variables.

Hi Karl

I tried 6.35.02.45 and that also gave high cpu load (~45%).  So it seems that 6.35.02.45 is the release note we should consider.

Is it possible that a default cache setting changed in 6.35.02.45 ?

Best regards

David

David,
what's the platform you are using? Is it some of the TI platforms delivered with XDCtools or you created your own? If the latter, can you attach the platform package?
Can you also attach your CFG script, just to see if there is anything interesting in there?

Sasha,

We created our own platform.  I have attached our CFG script.

Best regards

David

Sasha,

I have attached our package.

Best regards

David

Sasha,

Thanks for identifying the problem.  I have now specified L1D and L1P cache and the problem is fixed.

That was excellent support. Thank you!

BR

David