Software instabilities with EDK 10.01 and PPC405?!??!!!

My guess would be that you have a timing error. Does your design meet timing? Do you have a timing specification?

Are you running from external memory? Can you run a memory test on it?

Alan Nishioka

Hi,

That *could* be.

I have to dig into that on Monday....

Nope. I am using the EDK generated .ngc netlists and the EDK generated .ucf file. My design (it's connected via external memory controller) doesn't have critical timing and the synthesis estimate says 140Mhz.... So I didn't care for timing...

But what could I do if the EDK design does not meet timing specs?

Yes. DDR

Hm... I have tried the memory test demo application once. Perhaps I have to try it again....

Best regards,

Philipp :-)

Hello Philipp,

Are you using an operating system, or are you running stand alone code?

Your symptoms sound like they could be caused by a stack that is not big enough. If you are using an OS that supports virtual memory, the stack and heap sizes for each process are dealt with for you automatically. If not, than you need to pay some attention to make sure that their default values will work for your program.

printf can take a lot of memory resources, and every time you move printfs around to try and debug, the symptoms can change.

If you are running stand alone code, check what the stack and heap sizes are set to in EDK. In the Applications tab in the Project Information Area window, select your program, right click, then select Generate Linker Script. That will pop up a window that list the stack and heap sizes and other stuff. The default is 0x400 bytes, or 1k byte. That is big enough for our boot loader that loads Linux from a MiniSD card, and does some stuff with SPI roms. It uses xil_printf, but not libc printf.

Try increasing the stack and or heap sizes and see if that makes a difference.c

Regards,

John McCaskill

Hello John,

I am running standalone code. I am using the Xilinx supplied driver framework to get the CPU up and running etc.

No, currently no OS :-)

That's an important hint!

I just tried it - it MAKES a difference! But how can I determine the correct stack size? Stopping the program in the "deepest point", look at the stack pointer, add something for interrupt purposes and use that "counted" value? Which of my data goes to the heap? I don't use dynamic memory allocation (btw do the libxil and friends support a working malloc(), free() etc.?).

Many thanks for your tip :-)

There is a new EDK problem: I tried to change the PPC cpu speed by changing the cpu clock output frequency of the EDK clock generator. Did that via GUI and by editing MHS file. I can do a "make netlist" without any problems. But then, later, PAR gives me an error mentioning a dcm...clk..something signal that has a constraint but doesn't exist. I tried to find the constraint but I don't know where to look. I also grep'ed through the EDK generated netlists (after running ngc2edif) and did not find that signal. And I don't have a clue where I even could have a constraint file mentioning that signal....

Best wishes, Philipp :-)

Problem solved!

Ist was really a stack/heap insufficiency combined with a very little programming error (Forgot a wait condition related to the Xilinx IIC driver) that caused my design go mad. Now everything seems to be fine.

Thanks again,

Philipp :-)