effective wave to debug bsp_spurious_handler trap?

[Chan Kim]

 Hello, rtems users,
I'm seeing bsp_spurious_handler trap in an rtems appplication running on a sparc processor designed in our team.
It's kind of a test program. And it receives data from camera and performs convolution with weights read from SD card. Currently the trap occurs while convolution (many loops accessing memory) and I found when I put printf somewhere in the loop, I don't see the trap (but very slow). The number is printed sometimes 297 and sometimes 263. I tried printing the program counter and it is in printf and sometimes shows somewhere in a structure called 'rtems_filesystem_default_pathconf'. I'm not sure if this program counter is really where the trap happed. 
It should almost impossible to run hardware simulation for this complex program of course. Isn't there any well known trick that I can use debug where the trap is coming from? Of course I'll spend some time reading related documents.
Thanks.
Chan Kim / ETRI