When I unroll kernel function and use indirect address(1x4_10 or 1x4_9), Segment error (core dump)

[ysh329]

When I unroll kernel function (the third FOR LOOP in function) and use indirect address(1x4_10 or 1x4_9), build is okay.
however, it shows Segment error (core dump) when executing.

[ysh329]

I think and test my idea that this issue is relative to this code below:

for ( p = 0; p<k ; p+=4 )

there may exist a situation that p is bigger than k.

[ysh329]

This code doesn't implement border or special case.