What steps will reproduce the problem?
1. Generate a float/int array with #elements >= 64M;
2. Appropriately create the plan and call cudppScan on the array.
What is the expected output? What do you see instead?
Should do scan on the array. However, the behavior of the program is undefined.
What version of the product are you using? On what operating system?
1.1.1
Please provide any additional information below.
The current implementation only uses 1 dimensional grid, i.e. 64k blocks at
most, while each blocks process 1K elements. So it can at most process 64M
elements. Higher than this number will cause the kernel fail to launch. The
result returned is then undefined.
Original issue reported on code.google.com by likan...@gmail.com on 23 Jun 2010 at 1:27
Original issue reported on code.google.com by
likan...@gmail.com
on 23 Jun 2010 at 1:27