I'm experiencing this strange bug I have a kernel that iterates x elements over each other (just for information, x is a bit more than 3 million) like:
int i = get_global_id(0);
if (i >= Length[0]) return;
int count = Length[0];
for (int j = 0; j < count; j++)
I'm trying to get a simple test case, by doing a few more tests it looks it hangs when execution time exceeds one minute, more later.
I found that any kernel taking more than one min hangs no specific code, Windows 7 x64, HD5770, Cat 10.9
BTW, spliting my app in parts solved the problem, this would be needed anyway to run over multiple GPU but...
Please... Make opencl specific drivers that are not video drivers... Or at least a kill switch, pop corn and reset button are a painful workaround...
__kernel void test(__global int* Length, __global int* Count) { int j = 1; for(int i = 0; i < 3000000; i++) j *= Length[0] + i; Count[get_global_id(0)] = j; }
eduardoschardong,
I was not able to reproduce this issue on a vista machine on Juniper.
Have you tried catalyst 10.10.
On Catalyst 10.10 the watchdog timer kills after one minute...
Ok, all solved... But wasn't the watchdog timer supposed to kill after 5 minutes?