Maximum Number of Blocks?

[an_schall]

I want to start a kernel in such a fashion:

kernel_code<<<NUMBER_BLOCKS, NUMBER_THREADS_PER_BLOCK>>> (param1, param2, param3, param4);
Thus, using only the x-dimension of the grid. I want to call the kernel with the maximum number of blocks possible. I thought the max. number of blocks in a grid for one dimension would be 65535.

However, I explored the constant CU_DEVICE_ATTRIBUTE_MAX_GRID_DIM_X which sounds like exact the same number I want to find out. However, this constant returns 1899336 on my GeForce 210 (CUDA 1.2).

So my question is, what is the maximum number of blocks along one dimension for my GPU and how can I verify it?

Regards, P.

[pasoleatis]

You are mixing something for all devices with cc smaller than 3 it is 65536. In the nvidia samples in the 1-utilities folder there is a program call devicequery. Run it and it will give the mac number of blocks in x-dimension.

[an_schall]

Thanks a lot.