Hi,
I’m not sure if its the right place to inform about errors in documentation - anyway I haven’t found an email for error reporting.
In NVIDIA_OpenCL_BestPracticesGuide.pdf, page 26, Listing 3.9:
[codebox]
aTile[y][threadIdx.x] = a[row*TILE_DIM+x];
bTile[y][threadIdx.x] = b[y*N+col];
[/codebox]
There’s no threadIdx in OpenCL - apparently code samples were translated from CUDA. This makes me wonder if performance tests of code snippets in documentation were performed with OpenCL or were just taken from CUDA…