Would there be conflict if two threads access to the same data

[uboat]

If the blockis is 1616 and grid 31

and I let row = blockDim.x * blockIdx.x + threadIdx.x and no y coordinate. So if cuda calculates x[row], does it mean that all the threads in the same row all do this calculation? If so, different threads access to the same data, would there be any conflict?

And might there be the case that different blockDim.x blockIdx.x threadIdx.x but lead to the same value of row?

[lusine]

If the blockis is 1616 and grid 31

and I let row = blockDim.x * blockIdx.x + threadIdx.x and no y coordinate. So if cuda calculates x[row], does it mean that all the threads in the same row all do this calculation? If so, different threads access to the same data, would there be any conflict?

And might there be the case that different blockDim.x blockIdx.x threadIdx.x but lead to the same value of row?

Hello!..

I advice you to visit http://steps3d.narod.ru/tutorials/cuda-tutorial.html

There is an example of multiply matrices there.Different threads read same date but write in different positions.

When different threads write data in the same position it leads to the conflict.

Maybe your threads write data in the same position.