CUDA GPU SDK 4.0 has included a SCAN algorithm. However, the maximum array size is limited to 4*THREADBLOCK_SIZE * THREADBLOCK_SIZE, which is 262144. I tried to raise the limit, but got wrong results. Can somebody explain why the maximum array size is limited?
CUDA GPU SDK 4.0 has included a SCAN algorithm. However, the maximum array size is limited to 4*THREADBLOCK_SIZE * THREADBLOCK_SIZE, which is 262144. I tried to raise the limit, but got wrong results. Can somebody explain why the maximum array size is limited?